Zeitschriftenartikel zum Thema „TensorFlow Object Detection API 2“

Using the TensorFlow Object Detection API, an approach to identifying and registering Baikal diatom species Synedra acus subsp. radians has been tested. As a result, a set of images was formed and training was conducted. It is shown that аfter 15000 training iterations, the total value of the loss function was obtained equal to 0,04. At the same time, the classification accuracy is equal to 95%, and the accuracy of construction of the bounding box is also equal to 95%.

With the advancement of computing power of Smartphones, they seem to be a better option to be used as an Assistive Technology for the visually impaired. In this paper we have discussed an application which allows visually impaired users to detect objects of their choice in their environment. We have made use of the Tensorflow Lite Application Programmable Interface (API), an API by Tensorflow which specifically runs models on an Android Smartphone. We have discussed the architecture of the API and the application itself. We have discussed the performance of various types of models such as MobileNet, ResNet & Inception. We have compared the results of the various Models on their size, accuracy & inference time(ms) and found that the MobileNet has the best performance. We have also explained the working of our application in detail.

Detecting objects in real-time and converting them into an audio output was a challenging task. Recent advancement in computer vision has allowed the development of various real-time object detection applications. This paper describes a simple android app that would help the visually impaired people in understanding their surroundings. The information about the surrounding environment was captured through a phone’s camera where real-time object recognition through tensorflow’s object detection API was done. The detected objects were then converted into an audio output by using android’s text-to-speech library. Tensorflow lite made the offline processing of complex algorithms simple. The overall accuracy of the proposed system was found to be approximately 90%.

ABSTRAKPendeteksian golongan darah dilakukan untuk mengetahui golongan darah yang dimiliki. Hingga saat ini pendeteksian golongan darah masih dilakukan oleh petugas analis kesehatan menggunakan kemampuan mata manusia. Pada penelitian ini dilakukan perancangan alat pendeteksi golongan darah menggunakan ESP32-CAM. Alat ini menggunakan kamera OV2640 untuk menangkap citra, yang diproses menggunakan Tensorflow Object Detection API sebagai framework untuk melatih serta mengolah citra darah. Model latih akan digunakan pada kondisi pendeteksian langsung dan ditampilkan dalam bentuk jendela program golongan darah beserta tingkat akurasinya. Dalam penelitian ini pengujian dilakukan menggunakan 20 dataset dengan jarak pengukuran antara ESP32-CAM dengan citra golongan darah yaitu sejauh 20 cm. Hasil yang didapat selama pengujian mayoritas golongan darah yang dapat terdeteksi adalah golongan darah AB.Kata kunci: ESP32-CAM, Tensorflow, Python, Golongan Darah, Pengolahan Citra ABSTRACTBlood group detection is performed to determine the blood group. Currently, in detecting blood type, it still relies on the ability of the human eyeThis paper presents a human blood group detection device using ESP32-CAM. This tool uses ESP32-CAM to capture images, and the Tensorflow Object Detection API as a framework used to train and process an image. The way this tool works is that the ESP32-CAM will capture an image of the blood sample and then send it via the IP address. Through the IP Address, the python program will access the image, then the image will be processed based on a model that has been previously trained. The results of this processing will be displayed in the form of a window program along with the blood type and level of accuracy. In this study, testing was carried out based on the number of image samples, the number of datasets, and the measurement distance. The ideal measurement distance between the ESP32-CAM and the blood group image is 20 cm long. The results obtained during the testing of the majority of blood groups that can be detected are AB blood group.Keywords: ESP32-CAM, Tensorflow, Python, Blood Type, Image Processing

The aim of this research is to show the implementation of object detection on drone videos using TensorFlow object detection API. The function of the research is the recognition effect and performance of the popular target detection algorithm and feature extractor for recognizing people, trees, cars, and buildings from real-world video frames taken by drones. The study found that using different target detection algorithms on the “normal” image (an ordinary camera) has different performance effects on the number of instances, detection accuracy, and performance consumption of the target and the application of the algorithm to the image data acquired by the drone is different. Object detection is a key part of the realization of any robot’s complete autonomy, while unmanned aerial vehicles (UAVs) are a very active area of this field. In order to explore the performance of the most advanced target detection algorithm in the image data captured by UAV, we have done a lot of experiments to solve our functional problems and compared two different types of representative of the most advanced convolution target detection systems, such as SSD and Faster R-CNN, with MobileNet, GoogleNet/Inception, and ResNet50 base feature extractors.

<p>According to Global Adult Tobacco Survey 2016-17, 61.9% of people quitting tobacco the reason was the warnings displayed on the product covers. The focus of this paper is to automatically display warning messages in video clips. This paper explains the development of a system to automatically detect the smoking scenes using image recognition approach in video clips and then add the warning message to the viewer. The approach aims to detect the cigarette object using Tensorflow’s object detection API. Tensorflow is an open source software library for machine learning provided by Google which is broadly used in the field image recognition. At present, Faster R-CNN with Inception ResNet is theTensorflow’s slowest but most accurate model. Faster R-CNN with Inception Resnet v2 model is used to detect smoking scenes by training the model with cigarette as an object.</p><p><em><br /></em></p>

The Python Earth Engine application programming interface (API) provides a new open-source ecosphere for testing hydrocarbon detection algorithms on large volumes of images curated with the Google Earth Engine. We specifically demonstrate the Python Earth Engine API by calculating three hydrocarbon indices: fluorescence, rotation absorption, and normalized fluorescence. The Python Earth Engine API provides an ideal environment for testing these indices with varied oil seeps and spills by (1) removing barriers of proprietary software formats and (2) providing an extensive library of data analysis tools (e.g., Pandas and Seaborn) and classification algorithms (e.g., Scikit-learn and TensorFlow). Our results demonstrate end-member cases in which fluorescence and normalized fluorescence indices of seawater and oil are statistically similar and different. As expected, predictive classification is more effective and the calculated probability of oil is more accurate for scenarios in which seawater and oil are well separated in the fluorescence space.

Nowadays an educational mobile application has been widely accepted and opened new windows of opportunity to explore. With its flexibility and practicality, the mobile application can promote learning through playing with an interactive environment especially to the children. This paper describes the development of mobile learning to help children above 4 years old in learning English and Arabic language in a playful and fun way. The application is developed with a combination of Android Studio and the machine learning technique, TensorFlow object detection API in order to predict the output result. Developed application namely “LearnWithIman” has successfully been implemented and the results show the prediction of application is accurate based on the captured image with the list item. The inclusion of the user database for lesson tracking and new lesson will be added for improvement in the future.

The blind encounter commuting risks, such as failing to recognize and avoid obstacles while walking, but protective support systems are lacking. Acoustic signals at crosswalk lights are activated by button or remote control; however, these signals are difficult to operate and not always available (i.e., broken). Bollards are posts installed for pedestrian safety, but they can create dangerous situations in that the blind cannot see them. Therefore, we proposed an obstacle recognition system to assist the blind in walking safely outdoors; this system can recognize and guide the blind through two obstacles (crosswalk lights and bollards) with image training from the Google Object Detection application program interface (API) based on TensorFlow. The recognized results notify the blind through voice guidance playback in real time. The single shot multibox detector (SSD) MobileNet and faster region-convolutional neural network (R-CNN) models were applied to evaluate the obstacle recognition system; the latter model demonstrated better performance. Crosswalk lights were evaluated and found to perform better during the day than night. They were also analyzed to determine if a client could cross at a crosswalk, while the locations of bollards were analyzed by algorithms to guide the client by voice guidance.

This work explores the combination of free cloud computing, free open-source software, and deep learning methods to analyze a real, large-scale problem: the automatic country-wide identification and classification of surface mines and mining tailings dams in Brazil. Locations of officially registered mines and dams were obtained from the Brazilian government open data resource. Multispectral Sentinel-2 satellite imagery, obtained and processed at the Google Earth Engine platform, was used to train and test deep neural networks using the TensorFlow 2 application programming interface (API) and Google Colaboratory (Colab) platform. Fully convolutional neural networks were used in an innovative way to search for unregistered ore mines and tailing dams in large areas of the Brazilian territory. The efficacy of the approach is demonstrated by the discovery of 263 mines that do not have an official mining concession. This exploratory work highlights the potential of a set of new technologies, freely available, for the construction of low cost data science tools that have high social impact. At the same time, it discusses and seeks to suggest practical solutions for the complex and serious problem of illegal mining and the proliferation of tailings dams, which pose high risks to the population and the environment, especially in developing countries.

With the rapid progress of China’s urbanization, research on the automatic detection of land-use patterns in Chinese cities is of substantial importance. Deep learning is an effective method to extract image features. To take advantage of the deep-learning method in detecting urban land-use patterns, we applied a transfer-learning-based remote-sensing image approach to extract and classify features. Using the Google Tensorflow framework, a powerful convolution neural network (CNN) library was created. First, the transferred model was previously trained on ImageNet, one of the largest object-image data sets, to fully develop the model’s ability to generate feature vectors of standard remote-sensing land-cover data sets (UC Merced and WHU-SIRI). Then, a random-forest-based classifier was constructed and trained on these generated vectors to classify the actual urban land-use pattern on the scale of traffic analysis zones (TAZs). To avoid the multi-scale effect of remote-sensing imagery, a large random patch (LRP) method was used. The proposed method could efficiently obtain acceptable accuracy (OA = 0.794, Kappa = 0.737) for the study area. In addition, the results show that the proposed method can effectively overcome the multi-scale effect that occurs in urban land-use classification at the irregular land-parcel level. The proposed method can help planners monitor dynamic urban land use and evaluate the impact of urban-planning schemes.

In a soccer game the ability of humanoid robots that one needs to have is to see the ball object in real time. Development of the ability of humanoid robots to see the ball has been developed but the level of accuracy of object recognition and adaptation during matches still needs to be improved. The architecture designed in this study is Convolutional Neural Network or CNN which is designed to have 6 hidden layers with implementation of the robot program using the Tensorflow library. The pictures taken are used in the training process to have 9 types of images based on where the pictures were taken. Each type of image is divided into 2 classes, namely 2000 images for ball object classes and 2000 images for non-ball object classes. The test is done in real time using a white ball on green grass. From the architectural design and white ball detection test results obtained a success rate of 67%, five of the nine models managed to recognize the ball. The model can recognize objects with an image processing speed of a maximum of 13 FPS.Dalam pertandingan sepak bola kemampuan robot humanoid yang perlu dimiliki salah satunya adalah melihat objek bola secara real time. Pengembangan kemampuan robot humanoid untuk melihat bola telah dikembangkan tetapi tingkat akurasi pengenalan objek dan adaptasi saat pertandingan masih perlu ditingkatkan. Arsitektur yang dirancang pada penelitian ini yaitu Convolutional Neural Network atau CNN yang dirancang memiliki 6 hidden layer dengan implementasi pada program robot menggunakan library Tensorflow. Gambar yang diambil digunakan dalam proses training memiliki 9 jenis gambar berdasarkan tempat pengambilan gambar. Tiap jenis gambar terbagi menjadi 2 class yaitu 2000 gambar untuk class objek bola dan 2000 gambar untuk class objek bukan bola. Pengujian dilakukan secara real time dengan menggunakan bola berwarna putih di atas rumput hijau. Dari perancangan arsitektur dan hasil pengujian pendeteksian bola putih didapatkan persentase keberhasilan 67% yaitu lima dari sembilan model berhasil mengenali bola. Model dapat mengenali objek dengan kecepatan pengolahan gambar adalah maksimal 13 FPS.

<p><strong>Abstract.</strong> In this work we test the power of prediction of deep learning for detection of buildings from aerial laser scanner point cloud information. Automatic extraction of built features from remote sensing data is of extreme interest for many applications. In particular latest paradigms of 3D mapping of buildings, such as CityGML and BIM, can benefit from an initial determination of building geometries. In this work we used a LiDAR dataset of urban environment from the ISPRS benchmark on urban object detection. The dataset is labelled with eight classes, two were used for this investigation: roof and facades. The objective is to test how TensorFlow neural network for deep learning can predict these two classes. Results show that for “roof” and “facades” semantic classes respectively, recall is 84% and 76% and precision is 72% and 63%. The number and distribution of correct points well represent the geometry, thus allowing to use them as support for CityGML and BIM modelling. Further tuning of the hidden layers of the DL model will likely improve results and will be tested in future investigations.</p>

CAPTCHA authentication has been challenged by recent technology advances in AI. However, many of the AI advances challenging CAPTCHA are either restricted by a limited amount of labeled CAPTCHA data or are constructed in an expensive or complicated way. In contrast, this paper illustrates a low-cost approach that takes advantage of the nature of open source libraries for an AI-based chosen-plaintext attack. The chosen-plaintext attack described here relies on a deep learning model created and trained on a simple personal computer in a low-cost way. It shows an efficient cracking rate over two open-source Python CAPTCHA Libraries, Claptcha and Captcha. This chosen-plaintext attack method has raised a potential security alert in the era of AI, particularly to small-business owners who use the open-source CAPTCHA libraries. The main contributions of this project include: (1) it is the first low-cost method based on chosen-plaintext attack by using the nature of open-source Python CAPTCHA libraries; (2) it is a novel way to combine TensorFlow object detection and our proposed peak segmentation algorithm with convolutional neural network to improve the recognition accuracy.

Applications of science and technology have made a human life much easier. Vision plays a very important role in one’s life. Disease, accidents or due some other reasons people may loose their vision. Navigation becomes a major problem for the people with complete blindness or partial blindness. This paper aims to provide navigation guidance for visually impaired. Here we have designed a model which provides the instruction for the visionless people to navigate freely. NoIR camera is used to capture the picture around the person and identifies the objects. Using earphones voice output is provided defining the objects. This model includes Raspberry Pi 3 processor which collects the objects in surroundings and converts them into voice message, NoIR camera is used detect the object, power bank provides the power and earphones are used here the output message. TensorFlow API an open source software library used for object detection and classification. Using TensorFlow API multiple objects are obtained in a single frame. eSpeak a Text to Speech synthesizer (TTS) software is used to convert text (detected objects) to speech format. Hence using NoIR camera video which is captured is converted into voice output which provides the guidance for detecting objects. Using COCO model 90 commonly used objects are identified like person, table, book etc.

Autonomous or Self-driving vehicles are set to become the main mode of transportation for future generations. They are highly reliable, very safe and always improving as they never stop learning. There are numerous systems being developed currently based on various techniques like behavioural cloning and reinforcement learning. Almost all these systems work in a similar way, that is, the agent (vehicle) is completely aware of its immediate surroundings and takes future decisions based on its own historical experiences. The proposed work involves the design and implementation of Convolutional Neural Network (CNN) enhanced with new activation function. The proposed CNN is trained to take a picture of the road in front of it as input and give the required angle of tilt of the steering wheel . The model is trained using the behavioural cloning method and thus learns to navigate from the experiences of a human agent. This method is very accurate and efficient. In this paper, for the detection of object and vehicle in autonomous vehicle, the existing Tensorflow object Detection API is collaborated with pretrained SSD MobileNet model. This paper presents in detail literature survey on various techniques that have been used in predicting steering angle and object detection in self driving car. Apart from that, the effect of activation functions like ReLU, Sigmoid and ELU over the CNN model is analysed.

ObjectiveAutomated syndromic surveillance using mobile devices is an emerging public health focus that has a high potential for enhanced disease tracking and prevention in areas with poor infrastructure. Pacific Northwest National Laboratory sought to develop an Android mobile application for syndromic biosurveillance that would i) use the phone camera to take images of human faces to detect individuals that are sick through a machine learning (ML) model and ii) collect image data to increase training data available for ML models. The initial prototype use case is for screening and tracking the health of soldiers for use by the Department of Defense’s Disease Threat Reduction Agency.IntroductionInfectious diseases present with multifarious factors requiring several efforts to detect, prevent, and break the chain of transmission. Recently, machine learning has shown to be promising for automated surveillance leading to rapid and early interventions, and extraction of phenotypic features of human faces [3, 5]. In addition, mobile devices have become a promising tool to provide on-the-ground surveillance, especially in remote areas and geolocation mapping [4].Pacific Northwest National Laboratory (PNNL) combines machine learning with mobile technology to provide a groundbreaking prototype of disease surveillance without the need for internet, just a camera. In this android application, VisionDx, a machine learning algorithm analyses human face images and within milliseconds notifies the user with confidence level whether or not the person is sick. VisionDx comes with two modes, photo and video, and additional features of history, map, and statistics. This application is the first of its kind and provides a new way to think about the future of syndromic surveillance.MethodsData. Human healthy (n = 1096) and non-healthy (n = 1269) facial images met the criteria for training the Machine Learning model after preprocessing them. The healthy images were obtained from the Chicago face database [6] and California Institute of Technology [2]. There are no known collections of disease facial images. Using open source image collection/curation services, images were identified by a variety of keywords, including specific infectious diseases. The criteria for image inclusion was 1. a frontal face was identified using OpenCV library [1], and 2. the image contained signs of disease through visual inspection (e.g., abnormal color, texture, swelling).Model. To identify a sick face from a healthy one, we used transfer machine learning and experimented with various pretrained Convolutional Neural Networks (CNN) from Google for mobile and embedded vision applications. Using MobileNet, we trained the final model with our data and deployed it to our prototype mobile app. Google Mobile Vision API and TensorFlow mobile were used to detect human faces and run predictions in the mobile app.Mobile Application. The Android app was built using Android Studio to provide an easily navigable interface that connects every action between tabbed features. The app features (i.e., Map, Camera, History, and Statistics) are in tab view format. The custom-made camera is the main feature of the app, and it contains face detection capability. A real-time health status detection function gives a level of confidence based the algorithm results found on detected faces in the camera image.ResultsPNNL's prototype Android application, VisionDx, was built with user-friendly tab views and functions to take camera images of human faces and classify them as sick or healthy through an inbuilt ML model. The major functions of the app are the camera, map, history, and statistics pages. The camera tab has a custom-made camera with face detection algorithm and classification model of sick or healthy. The camera has image or video mode and results of the algorithm are updated in milliseconds. The Statistics view provides a simple pie chart on sick/healthy images based on user selected time and location. The Map shows pins representing all labeled images stored, and the History displays all the labeled images. Clicking on an image in either view shows the image with metadata, i.e., model confidence levels, geolocation, and datetime.The CNN model prediction accuracy has ~98% validation accuracy and ~96% test accuracy. High model performance shows the possibility that deep learning could be a powerful tool to detect sickness. However, given the limited dataset, this high accuracy also means the model is most likely overfit to the data. The training set is limited: a. the number of training images is small compared to the variability in facial expressions and skin coloring, and b. the sick images only contained overt clinical signs. If trained on a larger, diverse set of data, this prototype app could prove extremely useful in surveillance efforts of individual to large groups of people in remote areas, e.g., to identify individuals in need of medical attention or get an overview of population health. In effort to improve the model, VisionDx was developed as a data collection tool to build a more comprehensive dataset. Within the tool, users can override the model prediction, i.e., false positive or false negative, with a simple toggle button. Lastly, the app was built to protect privacy so that other phone aps can't access the images unless shared by a user.ConclusionsDeveloped at PNNL for the Defense Threat Reduction Agency, VisionDx is a novel, camera-based mobile application for real-time biosurveillance and early warning in the field without internet dependency. The prototype mobile app takes pictures of faces and analyzes them using a state-of-the-art machine learning model to give two confidence levels of likelihood of being sick and healthy. With further development of a labeled dataset, such as by using the app as a data collection too, the results of the algorithm will quickly improve leading to a ground-breaking approach to public health surveillance.References1. Bradski G. (n.d.) The OpenCV Library. Retrieved Sept 30, 2018 at http://www.drdobbs.com/open-source/the-opencv-library/1844043192. Computational Vision: Archive. (1999). Retrieved Sept 22, 2018 at http://www.vision.caltech.edu/html-files/archive.html3. Ferry Q, Steinberg J, Webber C, et al (2014). Diagnostically relevant facial gestalt information from ordinary photos. ELife, 3, e02020.4. Fornace KM, Surendra H, Abidin TR, et al (2018). Use of mobile technology-based participatory mapping approaches to geolocate health facility attendees for disease surveillance in low resource settings. International Journal of Health Geographics, 17(1), 21. https://doi.org/10.1186/s12942-018-0141-05. Lopez DM, de Mello FL, G Dias, CM, et al (2017). Evaluating the Surveillance System for Spotted Fever in Brazil Using Machine-Learning Techniques. Frontiers in Public Health, 5. https://doi.org/10.3389/fpubh.2017.003236. Ma DS, Correll J, Wittenbrink B. (2015) The Chicago face database: A free stimulus set of faces and norming data. Behavior Research Methods, 47(4), 1122–1135. https://doi.org/10.3758/s13428-014-0532-5