Zeitschriftenartikel zum Thema „Blurry frame detection“

Increasing grain production is essential to those areas where food is scarce. Increasing grain production by controlling crop diseases and pests in time should be effective. To construct video detection system for plant diseases and pests, and to build a real-time crop diseases and pests video detection system in the future, a deep learning-based video detection architecture with a custom backbone was proposed for detecting plant diseases and pests in videos. We first transformed the video into still frame, then sent the frame to the still-image detector for detection, and finally synthesized the frames into video. In the still-image detector, we used faster-RCNN as the framework. We used image-training models to detect relatively blurry videos. Additionally, a set of video-based evaluation metrics based on a machine learning classifier was proposed, which reflected the quality of video detection effectively in the experiments. Experiments showed that our system with the custom backbone was more suitable for detection of the untrained rice videos than VGG16, ResNet-50, ResNet-101 backbone system and YOLOv3 with our experimental environment.

Sometimes, model building in crystallography is like resolving a puzzle: All obvious symmetrical or methodological errors are excluded, you apparently understand the measured patterns in 3D, but the structure solution and/or refinement is just not working. One such nerve-stretching problem arises from metrically commensurate structures (MCS). This expression means that the observed values of the components of the modulation wave vectors are rational by chance and not because of a lock-in. Hence, it is not a superstructure - although the boundaries between the two descriptions are blurry. Using a superstructure model for a MCS decreases the degrees of freedom, and forces the atomic arrangement to an artificial state of ordering. Just imagine it as looking at a freeze frame from a movie instead of watching the whole film. The consequences in structure solution and refinement of MCS are not always as dramatically as stated in the beginning. On the contrary, treating a superstructure like a MCS might be a worthwhile idea. Converting from a superstructure model to a superspace model may lead to a substantial decrease in the number of parameters needed to model the structure. Further, it can permit for the refinement of parameters that the paucity of data does not allow in a conventional description. However, it is well known that families of superstructures can be described elegantly by the use of superspace models that collectively treat a whole range of structures, commensurate and incommensurate. Nevertheless, practical complications in the refinement are not uncommon. Instances are overlapping satellites from different orders and parameter correlations. Notably, MCS occur in intermetallic compounds that are important for the performance of next-generation electronic devices. Based on examples of their (pseudo)hexagonal 3+1D and 3+2D structures, we will discuss the detection and occurrence of MCS as well as the benefits and limitations of implementing them artificially.

Although there are well established object detection methods based on static images, their application to video data on a frame by frame basis faces two shortcomings: (i) lack of computational efficiency due to redundancy across image frames or by not using a temporal and spatial correlation of features across image frames, and (ii) lack of robustness to real-world conditions such as motion blur and occlusion. Since the introduction of the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) in 2015, a growing number of methods have appeared in the literature on video object detection, many of which have utilized deep learning models. The aim of this paper is to provide a review of these papers on video object detection. An overview of the existing datasets for video object detection together with commonly used evaluation metrics is first presented. Video object detection methods are then categorized and a description of each of them is stated. Two comparison tables are provided to see their differences in terms of both accuracy and computational efficiency. Finally, some future trends in video object detection to address the challenges involved are noted.

Pedestrian detection has attracted great research attention in video surveillance, traffic statistics, and especially in autonomous driving. To date, almost all pedestrian detection solutions are derived from conventional framed-based image sensors with limited reaction speed and high data redundancy. Dynamic vision sensor (DVS), which is inspired by biological retinas, efficiently captures the visual information with sparse, asynchronous events rather than dense, synchronous frames. It can eliminate redundant data transmission and avoid motion blur or data leakage in high-speed imaging applications. However, it is usually impractical to directly apply the event streams to conventional object detection algorithms. For this issue, we first propose a novel event-to-frame conversion method by integrating the inherent characteristics of events more efficiently. Moreover, we design an improved feature extraction network that can reuse intermediate features to further reduce the computational effort. We evaluate the performance of our proposed method on a custom dataset containing multiple real-world pedestrian scenes. The results indicate that our proposed method raised its pedestrian detection accuracy by about 5.6–10.8%, and its detection speed is nearly 20% faster than previously reported methods. Furthermore, it can achieve a processing speed of about 26 FPS and an AP of 87.43% when implanted on a single CPU so that it fully meets the requirement of real-time detection.

Current new developments in remote sensing imagery enable satellites to capture videos from space. These satellite videos record the motion of vehicles over a vast territory, offering significant advantages in traffic monitoring systems over ground-based systems. However, detecting vehicles in satellite videos are challenged by the low spatial resolution and the low contrast in each video frame. The vehicles in these videos are small, and most of them are blurred into their background regions. While region proposals are often generated for efficient target detection, they have limited performance on satellite videos. To meet this challenge, we propose a Local Region Proposing approach (LRP) with three steps in this study. A video frame is segmented into semantic regions first and possible targets are then detected in these coarse scale regions. A discrete Histogram Mixture Model (HistMM) is proposed in the third step to narrow down the region proposals by quantifying their likelihoods towards the target category, where the training is conducted on positive samples only. Experiment results demonstrate that LRP generates region proposals with improved target recall rates. When a slim Fast-RCNN detector is applied, LRP achieves better detection performance over the state-of-the-art approaches tested.

Today, the application of high-tech image format technology in contemporary visual experience and film and television production has become infinitely mature. In the era of the combination of modern technology and the Internet, virtual numbers connect the past with the future, merge reality, and myth and even synchronize the world between primitive and modern. This article adopts experimental analysis and comparative analysis, setting the experimental group and the reference group aim to use high-tech image formats from the perspective of a full-frame sensor to basically realize a perspective screen and a three-dimensional screen for observers in indoor scenes. In addition, the process of reconstructing a 3D model using high-precision geometric information and realistic color information is also described. The experimental results show that the sharpness threshold cannot be too small; otherwise, part of the clear image is misjudged as a blurred image. If the threshold is too large, the missed detection of the blurred image will increase. Combined with the subjective evaluation of the image, when the threshold is 0.8, the experimental result is close subjective evaluation, and the missed detection rate is 2.41%. This shows that the ASODVS three-dimensional digital scene constructed in this article can meet the needs of real-time image processing; it can effectively evaluate the clarity of realistic analog images. It shows that controlling the size of the y coordinate value can affect the user’s visual experience. The smaller the y value in a certain range, the clearer the result can be.

Most object detection models cannot achieve satisfactory performance under nighttime and other insufficient illumination conditions, which may be due to the collection of data sets and typical labeling conventions. Public data sets collected for object detection are usually photographed with sufficient ambient lighting. However, their labeling conventions typically focus on clear objects and ignore blurry and occluded objects. Consequently, the detection performance levels of traditional vehicle detection techniques are limited in nighttime environments without sufficient illumination. When objects occupy a small number of pixels and the existence of crucial features is infrequent, traditional convolutional neural networks (CNNs) may suffer from serious information loss due to the fixed number of convolutional operations. This study presents solutions for data collection and the labeling convention of nighttime data to handle various types of situations, including in-vehicle detection. Moreover, the study proposes a specifically optimized system based on the Faster region-based CNN model. The system has a processing speed of 16 frames per second for 500 × 375-pixel images, and it achieved a mean average precision (mAP) of 0.8497 in our validation segment involving urban nighttime and extremely inadequate lighting conditions. The experimental results demonstrated that our proposed methods can achieve high detection performance in various nighttime environments, such as urban nighttime conditions with insufficient illumination, and extremely dark conditions with nearly no lighting. The proposed system outperforms original methods that have an mAP value of approximately 0.2.

This study describes the adaptation of non-linear microscopy for single-particle tracking (SPT), a method commonly used in biology with single-photon fluorescence. Imaging moving objects with non-linear microscopy raises difficulties due to the scanning process of the acquisitions. The interest of the study is based on the balance between all the experimental parameters (objective, resolution, frame rate) which need to be optimized to record long trajectories with the best accuracy and frame rate. To evaluate the performance of the setup for SPT, several basic estimation methods are used and adapted to the new detection process. The covariance-based estimator (CVE) seems to be the best way to evaluate the diffusion coefficient from trajectories using the specific factors of motion blur and localization error.

ABSTRACT The rest-frame far to extreme ultraviolet (UV) colour–redshift relationship has been constructed from data on over $480\, 000$ quasars carefully cross-matched between SDSS Data Release 14 and the final GALEX photometric catalogue. UV matching and detection probabilities are given for all the quasars, including dependencies on separation, optical brightness, and redshift. Detection limits are also provided for all objects. The UV colour distributions are skewed redward at virtually all redshifts, especially when detection limits are accounted for. The median GALEX far-UV minus near-UV (FUV − NUV) colour–redshift relation is reliably determined up to z ≈ 2.8, corresponding to rest-frame wavelengths as short as 400 Å. Extreme UV (EUV) colours are substantially redder than found previously, when detection limits are properly accounted for. Quasar template spectra were forward modelled through the GALEX bandpasses, accounting for intergalactic opacity, intrinsic reddening, and continuum slope variations. Intergalactic absorption by itself cannot account for the very red EUV colours. The colour–redshift relation is consistent with no intrinsic reddening, at least for SMC-like extinction. The best model fit has a FUV continuum power-law slope αν, FUV = −0.34 ± 0.03 consistent with previous results, but an EUV slope αν, EUV = −2.90 ± 0.04 that is much redder and inconsistent with any previous composite value (all ≳ −2.0). The EUV slope difference can be attributed in part to the tendency of previous studies to preferentially select UV brighter and bluer objects. The weak EUV flux suggests quasar accretion disc models that include outflows such as disc winds.

We compare event-cameras with fast (global shutter) frame-cameras experimentally, asking: “What is the application domain, in which an event-camera surpasses a fast frame-camera?” Surprisingly, finding the answer has been difficult. Our methodology was to test event- and frame-cameras on generic computer vision tasks where event-camera advantages should manifest. We used two methods: (1) a controlled, cheap, and easily reproducible experiment (observing a marker on a rotating disk at varying speeds); (2) selecting one challenging practical ballistic experiment (observing a flying bullet having a ground truth provided by an ultra-high-speed expensive frame-camera). The experimental results include sampling/detection rates and position estimation errors as functions of illuminance and motion speed; and the minimum pixel latency of two commercial state-of-the-art event-cameras (ATIS, DVS240). Event-cameras respond more slowly to positive than to negative large and sudden contrast changes. They outperformed a frame-camera in bandwidth efficiency in all our experiments. Both camera types provide comparable position estimation accuracy. The better event-camera was limited by pixel latency when tracking small objects, resulting in motion blur effects. Sensor bandwidth limited the event-camera in object recognition. However, future generations of event-cameras might alleviate bandwidth limitations.

Intelligent Transportation Systems (ITS) is one of the most developing research topic along with growing advance technology and digital information. The benefits of research topic on ITS are to address some problems related to traffic conditions. Vehicle detection and tracking is one of the main step to realize the benefits of ITS. There are several problems related to vehicles detection and tracking. The appearance of shadow, illumination change, challenging weather, motion blur and dynamic background such a big challenges issue in vehicles detection and tracking. Vehicles detection in this paper using the Optical Flow Density algorithm by utilizing the gradient of object displacement on video frames. Gradient image feature and HSV color space on Optical flow density guarantee the object detection in illumination change and challenging weather for more robust accuracy. Hungarian Kalman filter algorithm used for vehicle tracking. Vehicle tracking used to solve miss detection problems caused by motion blur and dynamic background. Hungarian kalman filter combine the recursive state estimation and optimal solution assignment. The future positon estimation makes the vehicles detected although miss detection occurance on vehicles. Vehicles counting used single line counting after the vehicles pass that line. The average accuracy for each process of vehicles detection, tracking, and counting were 93.6%, 88.2% and 88.2% respectively.

The quality of multi-tag imaging greatly affects the effective detection of multi-tag. When multi-tag moves rapidly, the image may have serious dynamic blur, and tags can not be detected efficiently. In previous work, it is generally assumed that blur kernel and noise stationary to improve image quality. However, the dynamic deblurring of Radio Frequency Identification (RFID) multi-tag imaging is an ill-posed inverse problem. In this paper, firstly, blur-sharp multi-tag image pairs are made by superimposing and averaging the adjoin random frames. Then, we propose blind deblurring for dynamic RFID multi-tag imaging based on conditional generative adversarial nets (CGANs), which adds perceptual loss and content loss to generator to make image sharper. Finally, tags are detected by YOLOv3 in real time in end-to-end manner. Experimental results demonstrate that PSNR is at least 0.56dB higher and speed is at least 31.25 % faster than that of the current improved convolution neural networks (CNN). CGANs can remove image blur better, which has great superiority in the field of dynamic multi-tag imaging. In addition, YOLOv3 detects multi-tag quickly, thereby improving the detection accuracy.

Visual object tracking is an important research topic in the field of computer vision. Tracking–learning–detection (TLD) decomposes the tracking problem into three modules—tracking, learning, and detection—which provides effective ideas for solving the tracking problem. In order to improve the tracking performance of the TLD tracker, three improvements are proposed in this paper. The built-in tracking module is replaced with a kernelized correlation filter (KCF) algorithm based on the histogram of oriented gradient (HOG) descriptor in the tracking module. Failure detection is added for the response of KCF to identify whether KCF loses the target. A more specific detection area of the detection module is obtained through the estimated location provided by the tracking module. With the above operations, the scanning area of object detection is reduced, and a full frame search is required in the detection module if objects fails to be tracked in the tracking module. Comparative experiments were conducted on the object tracking benchmark (OTB) and the results showed that the tracking speed and accuracy was improved. Further, the TLD tracker performed better in different challenging scenarios with the proposed method, such as motion blur, occlusion, and environmental changes. Moreover, the improved TLD achieved outstanding tracking performance compared with common tracking algorithms.

In this paper, we propose an FPGA-based enhanced-SIFT with feature matching for stereo vision. Gaussian blur and difference of Gaussian pyramids are realized in parallel to accelerate the processing time required for multiple convolutions. As for the feature descriptor, a simple triangular identification approach with a look-up table is proposed to efficiently determine the direction and gradient of the feature points. Thus, the dimension of the feature descriptor in this paper is reduced by half compared to conventional approaches. As far as feature detection is concerned, the condition for high-contrast detection is simplified by moderately changing a threshold value, which also benefits the reduction of the resulting hardware in realization. The proposed enhanced-SIFT not only accelerates the operational speed but also reduces the hardware cost. The experiment results show that the proposed enhanced-SIFT reaches a frame rate of 205 fps for 640 × 480 images. Integrated with two enhanced-SIFT, a finite-area parallel checking is also proposed without the aid of external memory to improve the efficiency of feature matching. The resulting frame rate by the proposed stereo vision matching can be as high as 181 fps with good matching accuracy as demonstrated in the experimental results.

In this paper, text detection in moving business cards for helping visually impaired persons using a wearable camera is presented. The current methods to help visually impaired persons to read the natural scene text, menus and book covers have three problems. First, they assumed that the blind persons are standing still and the captured scene is not moving. Second, the blind persons do not know that the menu or the book cover is captured by the camera. Third, these methods cannot "see" business cards to help the blind persons. The proposed method includes moving detection, thumb detection motion blur detection and text detection methods. Experimental results show that the proposed method can reduce the time complexity. The reduced rates for training and testing sets are 83.18% and 77.08%, respectively. The text detection rates for training and testing sets are 93.44% and 94.58%, respectively. The fps is 53 for 320 × 240 video frames. The program size is 102KB and can be run on mobile devices.

Visual tracking requirement is increasing day by day due to the availability of high-performance digital cameras at low prices. Visual tracking becomes a complex problem when cameras suffer with unwanted and unintentional motion, resulting in motion-blurred unstabilized video. The problem in hand becomes more challenging when the target of interest is to be detected automatically in this unstabilized video. This paper presents a comprehensive single intelligent solution for these problems. The proposed algorithm auto-detects the camera motion, filters out the unintentional motion while stabilizing the video keeping intentional motion only using speeded-up robust features (SURF) technique. Motion smear due to unstabilization is also removed, providing sharp stabilized video output with video quality enhancement of up to 20dB. Gabor filter is used innovatively for auto-detection of target of interest in each stabilized frame. Then the target is tracked using SURF method.

Sign language is a way of expressing yourself with your body language, where every bit of ones expressions, goals, or sentiments are conveyed by physical practices, for example, outward appearances, body stance, motions, eye movements, touch and the utilization of space. Non-verbal communication exists in both creatures and people, yet this article concentrates on elucidations of human non-verbal or sign language interpretation into Hindi textual expression. The proposed method of implementation utilizes the image processing methods and synthetic intelligence strategies to get the goal of sign video recognition. To carry out the proposed task implementation it uses image processing methods such as frame analysing based tracking, edge detection, wavelet transform, erosion, dilation, blur elimination, noise elimination, on training videos. It also uses elliptical Fourier descriptors called SIFT for shape feature extraction and most important part analysis for feature set optimization and reduction. For result analysis, this paper uses different category videos such as sign of weeks, months, relations etc. Database of extracted outcomes are compared with the video fed to the system as a input of the signer by a trained unclear inference system.

Maize plant detection was conducted in this study with the goals of target fertilization and reduction of fertilization waste in weed spots and gaps between maize plants. The methods used included two types of color featuring and deep learning (DL). The four color indices used were excess green (ExG), excess red (ExR), ExG minus ExR, and the hue value from the HSV (hue, saturation, and value) color space, while the DL methods used were YOLOv3 and YOLOv3_tiny. For practical application, this study focused on performance comparison in detection accuracy, robustness to complex field conditions, and detection speed. Detection accuracy was evaluated by the resulting images, which were divided into three categories: true positive, false positive, and false negative. The robustness evaluation was performed by comparing the average intersection over union of each detection method across different sub–datasets—namely original subset, blur processing subset, increased brightness subset, and reduced brightness subset. The detection speed was evaluated by the indicator of frames per second. Results demonstrated that the DL methods outperformed the color index–based methods in detection accuracy and robustness to complex conditions, while they were inferior to color feature–based methods in detection speed. This research shows the application potential of deep learning technology in maize plant detection. Future efforts are needed to improve the detection speed for practical applications.

Multiwavelength light transmission imaging provides a possibility for early detection of breast cancer. However, due to strong scattering during the transmission process of breast tissue analysis, the transmitted image signal is weak and the image is blurred and this makes heterogeneous edge detection difficult. This paper proposes a method based on the weighted constraint decision (WCD) method to eliminate the erosion and checkerboard effects in image histogram equalization (HE) enhancement and to improve the recognition of heterogeneous edge. Multiwavelength transmission images of phantom are acquired on the designed experimental system and the mask image with high signal-to-noise ratio (SNR) is obtained by frame accumulation and an Otsu thresholding model. Then, during image enhancement the image is divided into low-gray-level (LGL) and high-gray-level (HGL) regions according to the distribution of light intensity in image. And the probability density distribution of gray level in the LGL and HGL regions are redefined respectively according to the WCD method. Finally, the reconstructed image is obtained based on the modified HE. The experimental results show that compared with traditional image enhancement methods, the WCD method proposed in this paper can greatly improve the contrast between heterogeneous region and normal region. Moreover, the correlation between the original image data is maintained to the greatest extent, so that the edge of the heterogeneity can be detected more accurately. In conclusion, the WCD method not only accurately identifies the edge of heterogeneity in multiwavelength transmission images, but it also could improve the clinical application of multiwavelength transmission images in the early detection of breast cancer.

AbstractIn this paper, the change detection of a fast turning specimen is studied at micro-level, whereas the images are acquired without stopping the rotation. In the beginning of the experiment, the imaging system is focused on the surface of the specimen. By starting the rotation of the specimen, the diameter of the specimen changes due to wear, which results in de-focusing of the imaging system. So the amount of blur in the images can be used as evidence of the wear phenomenon. Due to the properties of the microscope, the corners of the frames were dark and had to be cropped. So, each micrograph reflects only a small area of the surface. Nevertheless, techniques like stitching of multiple images can provide a significant surface area for micro-level investigation which increases the effectiveness of analyzing the material modification. Based on the results computer vision could detect a change of about 1.2 µm in the diameter of the specimen. More important is that we could follow the same locations of the surface in the microscopic images despite blurring, uneven illumination, change on the surface, and relatively a high-speed rotation.

A real-time detection and tracking method of landmark based for UAV visual navigation and fixed landing was proposed. This method used the SVM classification algorithm to train the offline classifier based on SURF-BoW features extracted from samples, which can recognize the landing landmark accurately and complete the initialization of the tracker. Afterwards, tracked the landmark via the improved median flow algorithm to ensure the reliability and integrity of the tracking target. Finally, based on the classifier and the principle of similarity between two adjacent frames' target, this paper designed a target re-search algorithm to ensure that the target can be retrieved quickly even if the target is lost or the target tracking fails, which makes the entire set of algorithm track the target accurately for a long time. The experimental results show that the proposed algorithm has good tracking performance under the conditions of the change of target scale, illumination changes and motion blur.

ABSTRACT We analyse 870 $\mu$m Atacama Large Millimetre Array (ALMA) dust continuum detections of 41 canonically selected $z$ ≃ 3 Lyman-break galaxies (LBGs), as well as 209 ALMA-undetected LBGs, in follow-up of SCUBA-2 mapping of the UKIDSS Ultra Deep Survey (UDS) field. We find that our ALMA-bright LBGs lie significantly off the local IRX-beta relation and have relatively bluer rest-frame UV slopes (as parametrized by β), given their high values of the ‘infrared excess’ (IRX ≡ LIR/LUV), relative to the average ‘local’ IRX-β relation. We attribute this finding in part to the young ages of the underlying stellar populations but we find that the main reason behind the unusually blue UV slopes are the relatively shallow slopes of the corresponding dust attenuation curves. We show that, when stellar masses, M*, are being established via SED fitting, it is absolutely crucial to allow the attenuation curves to vary (rather than fixing it on Calzetti-like law), where we find that the inappropriate curves may underestimate the resulting stellar masses by a factor of ≃2–3× on average. In addition, we find these LBGs to have relatively high specific star-formation rates (sSFRs), dominated by the dust component, as quantified via the fraction of obscured star formation $(f_{\rm obs}\equiv {\rm SFR_{\rm IR}/{\rm SFR}_{\rm UV+IR}})$. We conclude that the ALMA-bright LBGs are, by selection, massive galaxies undergoing a burst of a star formation (large sSFRs, driven, for example, by secular or merger processes), with a likely geometrical disconnection of the dust and stars, responsible for producing shallow dust attenuation curves.

Early-type galaxies (ETGs) are thought to be devoid of dust and star-formation, having formed most of their stars at early epochs. We present the detection of the dustiest ETGs in a large-area blind submillimetre survey with Herschel (H-ATLAS, Eales et al.2010), where the lack of pre-selection in other bands makes it the first unbiased survey for cold dust in ETGs. The parent sample of 1087 H-ATLAS galaxies in this study have a ⩾ 5σ detection at 250μm, a reliable optical counterpart to the submillimetre source (Smith et al.2011) and a spectroscopic redshift from the GAMA survey (Driver et al.2011). Additionally, we construct a control sample of 1052 optically selected galaxies undetected at 250μm and matched in stellar mass to the H-ATLAS parent sample to eliminate selection effects. ETGs were selected from both samples via visual classifications using SDSS images. Further details can be found in Rowlands et al. (2012). Physical parameters are derived for each galaxy using the multiwavelength spectral energy distribution (SED) fitting code of da Cunha, Charlot and Elbaz (2008), Smith et al.2012, using an energy balance argument. We investigate the differences between the dusty ETGs and the general ETG population, and find that the H-ATLAS ETGs are more than an order of magnitude dustier than the control ETGs. The mean dust mass of the 42 H-ATLAS ETGs is 5.5 × 107M⊙ (comparable to the dust mass of spirals in our sample), whereas the dust mass of the 233 control ETGs inferred from stacking at optical positions on the 250μm map is (0.8 - 4.0) × 106M⊙ for 25-15 K dust. The average star-formation rate of the H-ATLAS ETGs is 1.0 dex higher than that of control ETGs, and the mean r-band light-weighted age of the H-ATLAS ETGs is 1.8 Gyr younger than the control ETGs. The rest-frame NUV - r colours of the H-ATLAS ETGs are 1.0 magnitudes bluer than the control ETGs, and some ETGs may be transitioning from the blue cloud to the red sequence. Some H-ATLAS ETGs show signs of morphological disturbance and may have undergone recent rejuvenation of their ISM via gas and dust delivered by mergers. It is found that late-type stars cannot produce enough dust to account for that observed in the H-ATLAS ETGs. This indicates that either an external source of dust from mergers is required, a substantial amount of dust grain growth must occur in the ISM, or dust destruction by hot X-ray gas is less efficient than predicted.

In recent years, many agriculture-related problems have been evaluated with the integration of artificial intelligence techniques and remote sensing systems. Specifically, in fruit detection problems, several recent works were developed using Deep Learning (DL) methods applied in images acquired in different acquisition levels. However, the increasing use of anti-hail plastic net cover in commercial orchards highlights the importance of terrestrial remote sensing systems. Apples are one of the most highly-challenging fruits to be detected in images, mainly because of the target occlusion problem occurrence. Additionally, the introduction of high-density apple tree orchards makes the identification of single fruits a real challenge. To support farmers to detect apple fruits efficiently, this paper presents an approach based on the Adaptive Training Sample Selection (ATSS) deep learning method applied to close-range and low-cost terrestrial RGB images. The correct identification supports apple production forecasting and gives local producers a better idea of forthcoming management practices. The main advantage of the ATSS method is that only the center point of the objects is labeled, which is much more practicable and realistic than bounding-box annotations in heavily dense fruit orchards. Additionally, we evaluated other object detection methods such as RetinaNet, Libra Regions with Convolutional Neural Network (R-CNN), Cascade R-CNN, Faster R-CNN, Feature Selective Anchor-Free (FSAF), and High-Resolution Network (HRNet). The study area is a highly-dense apple orchard consisting of Fuji Suprema apple fruits (Malus domestica Borkh) located in a smallholder farm in the state of Santa Catarina (southern Brazil). A total of 398 terrestrial images were taken nearly perpendicularly in front of the trees by a professional camera, assuring both a good vertical coverage of the apple trees in terms of heights and overlapping between picture frames. After, the high-resolution RGB images were divided into several patches for helping the detection of small and/or occluded apples. A total of 3119, 840, and 2010 patches were used for training, validation, and testing, respectively. Moreover, the proposed method’s generalization capability was assessed by applying simulated image corruptions to the test set images with different severity levels, including noise, blurs, weather, and digital processing. Experiments were also conducted by varying the bounding box size (80, 100, 120, 140, 160, and 180 pixels) in the image original for the proposed approach. Our results showed that the ATSS-based method slightly outperformed all other deep learning methods, between 2.4% and 0.3%. Also, we verified that the best result was obtained with a bounding box size of 160 × 160 pixels. The proposed method was robust regarding most of the corruption, except for snow, frost, and fog weather conditions. Finally, a benchmark of the reported dataset is also generated and publicly available.

Background. Each 6 seconds 1 person in the world dies of a stroke. The stages of stroke care are the following: case detection, emergency medical care, diagnosis, treatment in the hyperacute and acute period, early and subsequent rehabilitation. The main symptoms of stroke can be identified with the help of FAST system (face – facial asymmetry, arm – limb weakness, speech – blurred speech, time – time after the event). Objective. To identify the main features of the management of stroke patients. Materials and methods. Analysis of the guidelines and literature data. Results and discussion. According to the recommendations of the American Heart Association for the emergency service, the time of processing of the call should be up to 90 seconds, the response time of the emergency service – 8 minutes, and the time spent at the place of care – up to 15 minutes. Upon admission to the hospital, the doctor of the stroke unit collects the anamnesis and takes a brief neurological examination, as well as determines the main physiological parameters. The nurse performs tests, the results of which come during thrombolysis. Then the patient undergoes a computed tomography, first aid is given and thrombolysis begins (still in the tomography apparatus). Time must be constantly documented. The median time should be 25 minutes. To organize the work of the stroke unit in the hyperacute and acute period of stroke it is necessary to create a local protocol – a document that lists the care procedures in a particular institution and describes the responsibility and coordination of individual units, time frames of each stage and basic medical services. It is also necessary to create a clinical roadmap for the patient, which includes a map of his moving, the content of procedures and their timer. Optimization of the stroke management is extremely important, as each 15 minutes of time saved leads to the 4 % reduction in mortality and to the 4 % increase in the likelihood of no adverse effects. In the first 3 hours it is especially important to monitor such parameters as body temperature, blood glucose, swallowing function (fever, sugar, swallowing – FESS) thoroughly. It is a significant problem, but for the most options of medical support the evidence base is insufficient, and therefore the doctor faces the choice to follow imperfect recommendations or to prescribe drugs according to the pathogenesis of cerebral ischemia. This choice has to be made in settings of severe time shortage, because the stroke kills 9,000 neurons every second. Stages of the ischemic cascade include ischemia, glutamate excitotoxicity, mitochondrial damage and dysfunction, oxidative and nitrosative stress, inflammation, and microcirculation disorders. Edaravone affects these stages, reducing the activity of oxygen radicals, the permeability of the blood-brain barrier, and the activity of secondary inflammation. This reduces the ischemic penumbra and the brain tissue necrosis area, which increases the likelihood of functional recovery. The study by M. Enomoto et al. (2019) showed that the use of edaravone and the endovascular reperfusion is a promising strategy for the treatment of ischemic stroke. Conclusions. 1. To coordinate the work of the stroke unit in the hyperacute and acute period of stroke, it is necessary to create local protocols and the clinical roadmap. 2. During the first 3 hours it is especially important to monitor such parameters as body temperature, blood glucose, and swallowing function. 3. Edaravone affects the ischemic stroke, reducing the activity of oxygen radicals, the permeability of the blood-brain barrier, and the activity of secondary inflammation.

In the current times, tasks like Object Detection, Object tracking, Gesture prediction, Video prediction in computer vision are being solved effectively with models of deep learning . Video frame prediction involves predicting the next few frames of a video given the previous frame or frames as input. Currently, the challenge in video frame prediction is that the predicted future frames are blurry. This paper focuses on the removal of noise from the predicted image using Denoising Autoencoders, solve the above-addressed issue. The proposed work, trains LSTM model which generates future frames by giving a sequence of input frames. The predicted output is given as an input to the Denoising Autoencoders which tries to remove the blurry predictions. Our approach is implemented on Moving MNIST Dataset. The result of our proposed method improved accuracy and is compared with the accuracy of Denoising Autoencoders, LSTM, and LSTM along with Denoising Autoencoders

Abstract Study question Can artificial intelligence (AI) algorithms identify spermatozoa in a semen sample without using training data annotated by professionals? Summary answer Unsupervised AI methods can discriminate the spermatozoon from other cells and debris. These unsupervised methods may have a potential for several applications in reproductive medicine. What is known already Identification of individual sperm is essential to assess a given sperm sample’s motility behaviour. Existing computer-aided systems need training data based on annotations by professionals, which is resource demanding. On the other hand, data analysed by unsupervised machine learning algorithms can improve supervised algorithms that are more stable for clinical applications. Therefore, unsupervised sperm identification can improve computer-aided sperm analysis systems predicting different aspects of sperm samples. Other possible applications are assessing kinematics and counting of spermatozoa. Study design, size, duration Three sperm-like paint images were manipulated using a graphic design tool and used to train our AI system. Two paintings have an ash colour background and randomly distributed white colour circles, and one painting has a predefined pattern of circles. Selected semen sample videos from a public dataset with videos obtained from 85 participants were used to test our AI system. Participants/materials, setting, methods Generative adversarial networks (GANs) have become common AI methods to process data in an unsupervised way. Based on single image frames extracted from videos, a GAN (SinGAN) can be trained to determine and track locations of sperms by translating the real images into localization paintings. The resulting model showed the potential of identifying the presence of sperms without any prior knowledge about data. Main results and the role of chance Visual comparisons of localization paintings to real sperm images show that inverse training of SinGANs can track sperms. Converting colour frames into grayscale frames and using grayscale synthetic sperm-like frames showed the best visual quality of generated localization paintings of sperm frames. Feeding real sperm video frames to the SinGAN at different scaling factors, which is defining the resolution of the input image, showed different quality levels of generated sperm localization paintings. A sperm frame given to the algorithm with a scaling factor of one leads to random sperm tracking, while the scales two to four result in more accurate localization maps than scaling levels five to eight. In contrast, scales from six to eight result in an output close to the input frame. The proposed method is robust in terms of the number of spermatozoa, meaning that the detection works well for samples with a low or high sperm count. For visual comparisons, visit our Github page: https://vlbthambawita.github.io/singan-sperm/. The sperm tracking speed of our SinGAN using an NVIDIA 1080 graphic processing unit, is around 17 frames per second, which can be improved by using parallel video processing capabilities. This shows the capability of using this method for real-time analysis. Limitations, reasons for caution Unsupervised methods are hard to train, and the results need human verification. The proposed method will need quality control and must be standardized. Unsupervised sperm tracking SinGAN may identify blurry bright spots as non-existing sperm heads which may restrict the use of SinGAN sperm tracking for sperm counting. Wider implications of the findings: Assessment of semen samples according to the WHO guidelines is subjective and resource-demanding. This unsupervised model might be used to develop new systems for less time-consuming and more accurate evaluation of semen samples. It may also be used for real-time analysis of prepared spermatozoa for use in assisted reproduction technology. Trial registration number N/A

The importance of efficient vehicle detection (VD) is increased with the expansion of road networks and the number of vehicles in the Intelligent Transportation Systems (ITS). This paper proposes a system for detecting vehicles at different weather conditions such as sunny, rainy, cloudy and foggy days. The first step to the proposed system implementation is to determine whether the video’s weather condition is normal or abnormal. The Random Forest (RF) weather condition classification was performed in the video while the features were extracted for the first two frames by using the Gray Level Co-occurrence Matrix (GLCM). In this system, the background subtraction was applied by the mixture of Gaussian 2 (MOG 2) then applying a number of pre-processing operations, such as Gaussian blur filter, dilation, erosion, and threshold. The main contribution of this paper is to propose a histogram equalization technique for complex weather conditions instead of a Gaussian blur filter for improving the video clarity, which leads to increase detection accuracy. Based on the previous steps, the system defines each region in the frame expected to contain vehicles. Finally, Support Vector Machine (SVM) classifies the defined regions into a vehicle or not. As compared to the previous methods, the proposed system showed high efficiency of about 96.4% and ability to detect vehicles at different weather conditions.

Abstract We present the first discovery of dust-reddened quasars (red quasars) in the high-z universe (z > 5.6). This is a result from the Subaru High-z Exploration of Low-Luminosity Quasars (SHELLQs) project, which is based on the sensitive multi-band optical imaging data produced by the Hyper Suprime-Cam (HSC) Subaru Strategic Program survey. We identified four red quasar candidates from the 93 spectroscopically confirmed high-z quasars in the SHELLQs sample, based on detections in the Wide-field Infrared Survey Explorer (WISE) data at 3.4 and 4.6 μm (rest-frame ∼5000–6500 Å). The amount of dust reddening was estimated with spectral energy distribution (SED) fits over optical and mid-infrared wavelengths. Two of the four candidates were found to be red quasars with dust reddening of E(B − V) > 0.1. The remaining SHELLQs quasars without individual WISE detections are significantly fainter in the WISE bands and bluer than the red quasars, although we did detect them in the W1 band in a stacked image. We also conducted the same SED fits for high-z optically-luminous quasars, but no red quasar was found. This demonstrates the power of Subaru HSC to discover high-z red quasars, which are fainter than the limiting magnitudes of past surveys in the rest-frame ultraviolet, due to dust extinction.

Event-based cameras are bio-inspired novel sensors that asynchronously record changes in illumination in the form of events. This principle results in significant advantages over conventional cameras, such as low power utilization, high dynamic range, and no motion blur. Moreover, by design, such cameras encode only the relative motion between the scene and the sensor and not the static background to yield a very sparse data structure. In this paper, we leverage these advantages of an event camera toward a critical vision application—video anomaly detection. We propose an anomaly detection solution in the event domain with a conditional Generative Adversarial Network (cGAN) made up of sparse submanifold convolution layers. Video analytics tasks such as anomaly detection depend on the motion history at each pixel. To enable this, we also put forward a generic unsupervised deep learning solution to learn a novel memory surface known as Deep Learning (DL) memory surface. DL memory surface encodes the temporal information readily available from these sensors while retaining the sparsity of event data. Since there is no existing dataset for anomaly detection in the event domain, we also provide an anomaly detection event dataset with a set of anomalies. We empirically validate our anomaly detection architecture, composed of sparse convolutional layers, on this proposed and online dataset. Careful analysis of the anomaly detection network reveals that the presented method results in a massive reduction in computational complexity with good performance compared to previous state-of-the-art conventional frame-based anomaly detection networks.

Sign language is one of the most popular language which is used as a communication bridge that depends on hand movement. As it is used worldwide, the variety of sign languages is very large and most of the people doesn’t understand it which makes the communication between deaf and normal people interrupted. This makes sign language recognition popular as it makes people doesn’t need to understand the sign but still understand the meaning of the sign. But sign language itself has many problems such as the possibility of different dataset has the same movement but different meaning, the method used of each dataset could be ineffective in other dataset, and many other else which makes it difficult to be implemented. Other than that, vision-based recognition is not as popular as sensor-based recognition because of the difference in feature accuracy even though it could give more area of improvement. That’s why the aim of this paper is to presents the combination of methods used to recognize vision-based Indonesian Sign Language and enhance the method using optimization technique. The methodology used in this study follows four steps framework of sign language recognition which is dataset collection, preprocessing, feature extraction, and recognition. Each method is improved in order to be compared and checked what is the method combination and optimization method is the best for sign language recognition. For dataset collection, the dataset that is used is formal Indonesian Sign Language which is called Sistem Isyarat Bahasa Indonesia (SIBI) with some constraint in order to make sure the quality of the dataset is good. The preprocessing methods are differentiated into two categories which is image enhancement and hand detection. Image enhancement methods include no image enhancement, Gaussian blur, Bilateral blur, and Contrast Limited Adaptive Histogram Equalization and hand detection methods include skin detection using YCbCr color space and edge detection using Canny algorithm. Feature that is used in this study is pair of left and right-side movement for each frame that is extracted by calculating the average value of each pixels from left and right-side image that could be a representative value of each frame. Lastly the recognition system that is used is Gaussian Hidden Markov Model and its’ state optimization which includes no optimization, Latin Hypercube optimization, Hill Climbing optimization, and Combination of Latin Hypercube and Hill Climbing optimization. The experiment result of the proposed method could recognize up to 82% accuracy rate. The improvement of this research could be implemented on

Halfway through the 2006 memoir comic Fun Home, the reader encounters a photograph that the book’s author, Alison Bechdel, found in a box of family snapshots shortly after her father’s death. The picture—“literally the core of the book, the centrefold” (Bechdel qtd. in Chute “Interview” 1006)—of Alison’s teenaged babysitter, Roy, erotically reclining on a bed in only his underwear, is the most tangible and direct evidence of her father’s sexual affairs with teenage boys, more confronting than his own earlier confession. Through this image, and a rich archive of familial texts, Bechdel chronicles her father’s thwarted desires and ambitions, probable suicide, and her own sexual and artistic coming of age.Bruce Bechdel, a married school teacher and part-time funeral director, was also an avid amateur historical restorer and connoisseur of modernist literature. Shortly after Alison came out to her parents at nineteen, Bruce was hit by a truck in what his daughter believes was an act of suicide. In Fun Home, Bechdel reads her family history suspiciously, plumbing family snapshots, letters, and favoured novels, interpreting against the grain, to trace her queer genealogy. Ultimately, she inverts this suspicious and interrogative reading, using the evidence she has gathered in order to read her father’s sexuality positively and embrace her queer and artistic inheritance from him. In The New York Times Magazine, in 2004, Charles McGrath made the suggestion that comics were “the new literary form” (24). Although comics have not yet reached widespread mainstream acceptance as a medium of merit, the burgeoning field of comics scholarship over the last fifteen years, the 2007 adaptation of Marjane Satrapi’s Persepolis into a feature film, and the addition of comics to the Best American series all testify to the widening popularity and status of the form. Memoir comics have established themselves, as Hillary Chute notes, as “the dominant mode of contemporary work” (Graphic 17). Many of these autobiographical works, including Fun Home, recount traumatic histories, employing the medium’s unique capacity to evoke the fractured and repetitive experience of the traumatised through panel structure and use of images. Comics articulate “what wasn’t permitted to be said or imagined, defying the ordinary processes of thought” (Said qtd. in Whitlock 967). The hand-drawn nature of comics emphasises the subjectivity of perception and memory, making it a particularly powerful medium for personal histories. The clear mediation of a history by the artist’s hand complicates truth claims. Comics open up avenues for both suspicious and restorative readings because their form suggests that history is always constructed and therefore not able to be confirmed as “ultimately truthful,” but also that there is no ultimate truth to be unveiled. No narrative is unmediated; a timeline is not more “pure” than a fleshed out narrative text. All narratives exclude information in order to craft a comprehensible series of events. Bechdel’s role as a suspicious reader of her father and of her own history resonates through her role as a historian and her interrogation of the ethical concerns of referential writing.Eve Kosofsky Sedgwick’s Touching Feeling: Affect, Pedagogy, Performativity critiques the hermeneutics of suspicion from a queer theory perspective, instead advocating reparative reading as a critical strategy. The hermeneutics of suspicion describes “the well-oiled machine of ideology critique” that has become the primary mode of critical reading over the last thirty or so years, suspiciously interpreting texts to uncover their hidden ideological biases (Felski, Uses 1). Reparative reading, on the other hand, moves away from this paranoid mode, instead valuing pleasure and “positive affects like joy and excitement” (Vincent). Sedgwick does not wholly reject suspicious reading, suggesting that it “represent[s] a way, among other ways, of seeking, finding, and organizing knowledge. Paranoia knows some things well and others poorly” (Touching 129). Felski, paraphrasing Ricoeur, notes that the hermeneutics of suspicion “adopts an adversarial sensibility to probe for concealed, repressed, or disavowed meanings” (“Suspicious” 216). In this fashion, Bechdel employs suspicious strategies to reveal her father’s hidden desires and transgressions that were obscured in the standard version of her family narrative, but ultimately moves away from such techniques to joyfully embrace her inheritance from him. Sedgwick notes that paranoid readings may only reveal that which is already known:While there is plenty of hidden violence that requires exposure there is also, and increasingly, an ethos where forms of violence that are hypervisible from the start may be offered as an exemplary spectacle rather than remain to be unveiled as a scandalous secret. (Touching 139)This is contrary to suspicious reading’s assumption that violence is culturally shunned, hidden, and in need of “unveiling” in contemporary Western culture. It would be too obvious for Bechdel to condemn her father: gay men have been unfairly misrepresented in the American popular imagination for decades, if not longer. Through her reparative reading of him, she rejects this single-minded reduction of people to one negative type. She accepts both her father’s weaknesses and her debts to him. A reading which only sought to publicise Bruce’s homosexual affairs would lack the great depth that Bechdel finds in the slippage between her father’s identity and her own.Bechdel’s embrace of Bruce’s failings as a father, a husband, and an artist, her revisioning of his death as a positive, creative act full of agency, and her characterisation of him as a supportive forerunner, “there to catch [Alison] as [she] leapt,” (Bechdel 232) moves his story away from archetypal narratives of homosexual tragedy. Bechdel’s memoir ends with (and enacts through its virtuoso execution) her own success, and the support of those who came before her. This move mirrors Joseph Litvak’s suggestion that “the importance of ‘mistakes’ in queer reading and writing […] has a lot to do with loosening the traumatic, inevitable-seeming connection between mistakes and humiliation […] Doesn’t reading queer mean learning, among other things, that mistakes can be good rather than bad surprises?” (Sedgwick Touching 146–7).Fun Home is saturated with intertextual references and archival materials that attempt to piece together the memoir’s fractured and hidden histories. The construction of this personal history works by including familial and historical records to register the trauma of the Bechdels’ personal tragedy. The archival texts are meticulously hand-drawn, their time-worn and ragged physicality maintained to emphasise the referentiality of these documents. Bechdel’s use of realistically drawn family photographs, complete with photo corners, suggests a family photograph album, although rather than establishing a censored and idealistic narrative, as most family albums do, the photographs are read and reproduced for their suppressed and destabilising content. Bechdel describes them as “particularly mythic” (Chute “Interview” 1009), and she plunders this symbolic richness to rewrite her family history. The archival documents function as primary texts, which stand in opposition to the deadly secrecy of her childhood home: they are concrete and evidentiary. Bechdel reads her father’s letters and photographs (and their gothic revival house) for sexual and artistic evidence, “read[ing] the text against the grain in order to draw out what it refuses to own up to” (Felski “Suspicious” 23). She interprets his letters’ baroque lyrical flourishes as indications both of his semi-repressed homosexuality and of the artistic sensibility that she would inherit and refine.Suspicion of the entire historical project marks the memoir. Philippe Lejeune describes the “Autobiographical Pact” as “a contract of identity that is sealed by the proper name” of the author (19). Bechdel does not challenge this pact fundamentally—the authoritative narrative voice of her book structures it to be read as historically truthful—but she does challenge and complicate the apparent simplicity of this referential model. Bechdel’s discussion of the referential failings of her childhood diary making—“the troubled gap between word and meaning”—casts a suspicious eye over the rest of the memoir’s historical project (Bechdel 143). She asks how language can adequately articulate experience or refer to the external world in an environment defined by secrets and silence. At the time of her childhood, it cannot—the claim to full disclosure that the memoir ultimately makes is predicated on distance and time. Bechdel simultaneously makes a claim for the historical veracity of her narrative and destabilises our assumptions around the idea of factual and retrospective truth:When I was ten, I was obsessed with making sure my diary entries bore no false witness. But as I aged, hard facts gave way to vagaries of emotion and opinion. False humility, overwrought penmanship, and self-disgust began to cloud my testimony […] until […] the truth is barely perceptible behind a hedge of qualifiers, encryption, and stray punctuation. (Bechdel 169)That which is “unrepresentable” is simultaneously represented and denied. The comics medium itself, with its simultaneous graphic and textual representation, suggests the unreliability of any one means of representation. Of Bechdel’s diaries, Jared Gardner notes, “what develops over the course of her diary […] is an increasing sense that text and image are each alone inadequate to the task, and that some merger of the two is required to tell the story of the truth, and the truth of the story” (“Archives” 3).As the boyishly dressed Alison urges her father, applying scare-quoted “bronzer,” to hurry up, Bechdel narrates, “my father began to seem morally suspect to me long before I knew that he actually had a dark secret” (16). Alison is presented as her father’s binary opposite, “butch to his nelly. Utilitarian to his aesthete,” (15) and, as a teenager, frames his love of art and extravagance as debauched. This clear distinction soon becomes blurred, as Alison and Bruce’s similarities begin to overwhelm their differences. The huge drawn hand shown holding the photograph of Roy, in the memoir’s “centrefold,” more than twice life-size, reproduces the reader’s hand holding the book. We are placed in Bechdel’s, and by extension her father’s, role, as the illicit and transgressive voyeurs of the erotic spectacle of Roy’s body, and as the possessors and consumers of hidden, troubling texts. At this point, Bechdel begins to take her queer reading of this family archive and use it to establish a strong connection between her initially unsympathetic father and herself. Despite his neglect of his children, and his self-involvement, Bechdel claims him as her spiritual and creative father, as well as her biological one. This reparative embrace moves Bruce from the role of criticised outsider in Alison’s world to one of queer predecessor. Bechdel figures herself and her father as doubled aesthetic and erotic observers and appreciators. Ann Cvetkovich suggests that “mimicking her father as witness to the image, Alison is brought closer to him only at the risk of replicating his illicit sexual desires” (118). For Alison, consuming her father’s texts connects her with him in a positive yet troubling way: “My father’s end was my beginning. Or more precisely, […] the end of his lie coincided with the beginning of my truth” (Bechdel 116–17). The final panel of the same chapter depicts Alison’s hands holding drawn photos of herself at twenty-one and Bruce at twenty-two. The snapshots overlap, and Bechdel lists the similarities between the photographs, concluding, “it’s about as close as a translation can get” (120). Through the “vast network of transversals” (102) that is their life together, Alison and Bruce are, paradoxically, twinned “inversions of one another” (98). Sedgwick suggests that “inversion models […] locate gay people—whether biologically or culturally—at the threshold between genders” (Epistemology 88). Bechdel’s focus on Proust’s “antiquated clinical term” both neatly fits her thematic expression of Alison and Bruce’s relationship as doubles (“Not only were we inverts. We were inversions of one another”) and situates them in a space of possibility and liminality (97-98).Bechdel rejects a wholly suspicious approach by maintaining and embracing the aporia in her and her father’s story, an essential element of memory. According to Chute, Fun Home shows “that the form of comics crucially retains the insolvable gaps of family history” (Graphic 175). Rejecting suspicion involves embracing ambiguity and unresolvability. It concedes that there is no one authentic truth to be neatly revealed and resolved. Fun Home’s “spatial and semantic gaps […] express a critical unknowability or undecidability” (Chute Graphic 182). Bechdel allows the gaps in her narrative to remain, refusing to “pretend to know” Bruce’s “erotic truth” (230), an act to which suspicious reading is diametrically opposed. Suspicious reading wishes to close all gaps, to articulate silences and literalise mysteries, and Bechdel’s narrative progressively moves away from this mode. The medium of comics uses words and images together, simultaneously separate and united. Similarly, Alison and Bruce are presented as opposites: butch/sissy, artist/dilettante. Yet the memoir’s conclusion presents Alison and Bruce in a loving, reciprocal relationship. The final page of the book has two frames: one of Bruce’s perspective in the moment before his death, and one showing him contentedly playing with a young Alison in a swimming pool—death contrasted with life. The gaps in the narrative are not closed but embraced. Bechdel’s “tricky reverse narration” (232) suggests a complex mode of reading that allows both Bechdel and the reader to perceive Bruce as a positive forebear. Comics as a medium pay particular visual attention to absence and silence. The gutter, the space between panels, functions in a way that is not quite paralleled by silence in speech and music, and spaces and line breaks in text—after all, there are still blank spaces between words and elements of the image within the comics panel. The gutter is the space where closure occurs, allowing readers to infer causality and often the passing of time (McCloud 5). The gutters in this book echo the many gaps in knowledge and presence that mark the narrative. Fun Home is impelled by absence on a practical level: the absence of the dead parent, the absence of a past that was unspoken of and yet informed every element of Alison’s childhood.Bechdel’s hyper-literate narration steers the reader through the memoir and acknowledges its own aporia. Fun Home “does not seek to preserve the past as it was, as its archival obsession might suggest, but rather to circulate ideas about the past with gaps fully intact” (Chute Graphic 180). Bechdel, while making her own interpretation of her father’s death clear, does not insist on her reading. While Bruce attempted to restore his home into a perfect, hermetically sealed simulacrum of nineteenth-century domestic glamour, Bechdel creates a postmodern text that slips easily between a multiplicity of time periods, opening up the absences, failures, and humiliations of her story. Chute argues:Bruce Bechdel wants the past to be whole; Alison Bechdel makes it free-floating […] She animates the past in a book that is […] a counterarchitecture to the stifling, shame-filled house in which she grew up: she animates and releases its histories, circulating them and giving them life even when they devolve on death. (Graphic 216)Bechdel employs a literary process of detection in the revelation of both of their sexualities. Her archive is constructed like an evidence file; through layered tableaux of letters, novels and photographs, we see how Bruce’s obsessive love of avant-garde literature functions as an emblem of his hidden desire; Alison discovers her sexuality through the memoirs of Colette and the seminal gay pride manifestos of the late 1970s. Watson suggests that the “panels, gutters, and page, as bounded and delimited visual space, allow texturing of the two-dimensional image through collage, counterpoint, the superimposition of multiple media, and self-referential gestures […] Bechdel's rich exploitation of visual possibilities places Fun Home at an autobiographical interface where disparate modes of self-inscription intersect and comment upon one another” (32).Alison’s role as a literary and literal detective of concealed sexualities and of texts is particularly evident in the scene when she realises that she is gay. Wearing a plaid trench coat with the collar turned up like a private eye, she stands in the campus bookshop reading a copy of Word is Out, with a shadowy figure in the background (one whose silhouette resembles her father’s teenaged lover, Roy), and a speech bubble with a single exclamation mark articulating her realisation. While “the classic detective novel […] depends on […] a double plot, telling the story of a crime via the story of its investigation” (Felski “Suspicious” 225), Fun Home tells the story of Alison’s coming out and genesis as an artist through the story of her father’s brief life and thwarted desires. On the memoir’s final page, revisioning the artifactual photograph that begins her final chapter, Bechdel reclaims her father from what a cool reading of the historical record (adultery with adolescents, verbally abusive, emotionally distant) might encourage readers to superficially assume. Cvetkovich articulates the way Fun Home uses:Ordinary experience as an opening onto revisionist histories that avoid the emotional simplifications that can sometimes accompany representations of even the most unassimilable historical traumas […] Bechdel refuses easy distinctions between heroes and perpetrators, but doing so via a figure who represents a highly stigmatised sexuality is a bold move. (125)Rejecting paranoid strategies, Bechdel is less interested in classification and condemnation of her father than she is in her own tangled relation to him. She adopts a reparative strategy by focusing on the strands of joy and identification in her history with her father, rather than simply making a paranoid attack on his character.She occludes the negative possibilities and connotations of her father’s story to end on a largely positive note: “But in the tricky reverse narration that impels our entwined stories, he was there to catch me when I leapt” (232). In the final moment of her text Bechdel moves away from the memoir’s earlier destabilising actions, which forced the reader to regard Bruce with suspicion, as the keeper of destructive secrets and as a menacing presence in the Bechdels’ family life. The final image is of complete trust and support. His death is rendered not as chaotic and violent as it historically was, but calm, controlled, beneficent. Bechdel has commented, “I think it’s part of my father’s brilliance, the fact that his death was so ambiguous […] The idea that he could pull that off. That it was his last great wheeze. I want to believe that he went out triumphantly” (qtd. in Burkeman). The revisioning of Bruce’s death as a suicide and the reverse narration which establishes the accomplished artist and writer Bechdel’s creative and literary debt to him function as a redemption.Bechdel queers her suspicious reading of her family history in order to reparatively reclaim her father’s historical and personal connection with herself. The narrative testifies to Bruce’s failings as a father and husband, and confesses to Alison’s own complicity in her father’s transgressive desires and artistic interest, and to her inability to represent the past authoritatively and with complete accuracy. Bechdel both engages in and ultimately rejects a suspicious interpretation of her family and personal history. As Gardner notes, “only by allowing the past to bleed into history, fact to bleed into fiction, image into text, might we begin to allow our own pain to bleed into the other, and more urgently, the pain of the other to bleed into ourselves” (“Autobiography’s” 23). Suspicion itself is queered in the reparative revisioning of Bruce’s life and death, and in the “tricky reverse narration” (232) of the künstlerroman’s joyful conclusion.ReferencesBechdel, Alison. Fun Home: A Family Tragicomic. New York: Mariner Books, 2007. Burkeman, Oliver. “A life stripped bare.” The Guardian 16 Oct. 2006: G2 16.Cvetkovich, Ann. “Drawing the Archive in Alison Bechdel’s Fun Home.” Women’s Studies Quarterly 36.1/2 (2008): 111–29. Chute, Hillary L. Graphic Women: Life Narrative and Contemporary Comics. New York: Columbia UP, 2010. ---. “Interview with Alison Bechdel.” MFS Modern Fiction Studies 52.4 (2006): 1004–13. Felski, Rita. Uses of Literature. Malden: Blackwell Publishing, 2008.---. “Suspicious Minds.” Poetics Today 32:3 (2011): 215–34. Gardner, Jared. “Archives, Collectors, and the New Media Work of Comics.” MFS Modern Fiction Studies 52.4 (2006): 787–806. ---. “Autobiography’s Biography 1972-2007.” Biography 31.1 (2008): 1–26. Lejeune, Philippe. On Autobiography. Ed. Paul John Eakin. Trans. Katherine Leary. Minneapolis: University of Minnesota Press, 1989. McCloud, Scott. Understanding Comics: The Invisible Art. New York: HarperPerennial, 1994. McGrath, Charles. “Not Funnies.” New York Times Magazine 11 Jul. 2004: 24–56. Sedgwick, Eve Kosofsky. Epistemology of the Closet. Berkeley: University of California Press, 2008. ---. Touching Feeling. Durham : Duke University Press, 2003. Vincent, J. Keith. “Affect and Reparative Reading.” Honoring Eve. Ed. J. Keith Vincent. Affect and Reparative Reading. Boston University College of Arts and Sciences. October 31 2009. 25 May 2011. ‹http://www.bu.edu/honoringeve/panels/affect-and-reparative-reading/?›.Watson, Julia. “Autographic disclosures and genealogies of desire in Alison Bechdel’s Fun Home.” Biography 31.1 (2008): 27–59. Whitlock, Gillian. “Autographics: The Seeing “I” of the Comics.” Modern Fiction Studies 52.4 (2006): 965–79.