141 |
A Transfer Learning Approach to Object Detection Acceleration for Embedded ApplicationsVance, Lauren M. 08 1900 (has links)
Indiana University-Purdue University Indianapolis (IUPUI) / Deep learning solutions to computer vision tasks have revolutionized many industries in recent years, but embedded systems have too many restrictions to take advantage of current state-of-the-art configurations. Typical embedded processor hardware configurations must meet very low power and memory constraints to maintain small and lightweight packaging, and the architectures of the current best deep learning models are too computationally-intensive for these hardware configurations. Current research shows that convolutional neural networks (CNNs) can be deployed with a few architectural modifications on Field-Programmable Gate Arrays (FPGAs) resulting in minimal loss of accuracy, similar or decreased processing speeds, and lower power consumption when compared to general-purpose Central Processing Units (CPUs) and Graphics Processing Units (GPUs). This research contributes further to these findings with the FPGA implementation of a YOLOv4 object detection model that was developed with the use of transfer learning. The transfer-learned model uses the weights of a model pre-trained on the MS-COCO dataset as a starting point then fine-tunes only the output layers for detection on more specific objects of five classes. The model architecture was then modified slightly for compatibility with the FPGA hardware using techniques such as weight quantization and replacing unsupported activation layer types. The model was deployed on three different hardware setups (CPU, GPU, FPGA) for inference on a test set of 100 images. It was found that the FPGA was able to achieve real-time inference speeds of 33.77 frames-per-second, a speedup of 7.74 frames-per-second when compared to GPU deployment. The model also consumed 96% less power than a GPU configuration with only approximately 4% average loss in accuracy across all 5 classes. The results are even more striking when compared to CPU deployment, with 131.7-times speedup in inference throughput. CPUs have long since been outperformed by GPUs for deep learning applications but are used in most embedded systems. These results further illustrate the advantages of FPGAs for deep learning inference on embedded systems even when transfer learning is used for an efficient end-to-end deployment process. This work advances current state-of-the-art with the implementation of a YOLOv4 object detection model developed with transfer learning for FPGA deployment.
|
142 |
Modeling of Video Quality for Automatic Video Analysis and Its Applications in Wireless Camera NetworksKong, Lingchao 01 October 2019 (has links)
No description available.
|
143 |
Detekce rychlosti přibližování automobilu na základě zpracování obrazu kamery / Detection of car approach speed using camera imge processingKovář, Jan January 2010 (has links)
The thesis deals with digital image processing, from the initial acquisition of digital picture frames, subsequent processing segmentation and algorithms to detect visual shapes on the scene. Image processing is a very broad topic, so here are analyzed for more understanding of the fundamental principles of perception and processing of video signals, image representation, his starting shooting through filters governing digital image processing methods to detect the objects in an image. It is also demonstrated by the size dependence of the object in the image on the distance from the camera, whereby we can determine the speed of approaching or moving away from the object. We will show you the specific determination of the distance we need to know the actual result size of the object. This is because the ratio between the size of the object depending on the distance is the same for each object. Finally, this work presents the resulting image frames for implementation using OpenCV library.
|
144 |
Sledování objektu ve videosekvenci pomocí integrálního histogramu / Object tracking in video sequence using the integral histogramPřibyl, Jakub January 2020 (has links)
This thesis focuses on object tracking in real-time. Tracked object is defined by bounding rectangle. The thesis works on issue of image processing and using histogram for real-time object tracking. The main contribution of the work is the extension of the provided program to track object in real-time with changing bounding rectangle. Size of the rectangle is changing as the object moves closer of further from camera. Furthemore the detection behavior in different scenarios is analyzed. In addition, various weight calculations were tested. The program is written in C++ using OpenCV library.
|
145 |
Detection of Aircraft, Vehicles and Ships in Aerial and Satellite Imagery using Evolutionary Deep LearningThoudoju, Akshay Kumar January 2021 (has links)
Background. The view of the Earth from above can offer a lot of data and with technological advancements in image sensors and high-resolution satellite images there is more quantity and quality of data which can be useful in research and applications like military, monitoring climate, etc. Deep neural networks have been successful in object detection and it is seen that their learning process can be improved with using right hyperparameters when configuring the networks. This can be done hyperparameter optimization by the use of genetic algorithms. Objectives. The thesis focuses on obtaining deep learning techniques with optimal hyperparameters using genetic algorithm to detect aircraft, vehicles and ships from satellite and aerial images and compare the optimal models with the original deep learning models. Methods. The study uses literature review to obtain the appropriate deep learning techniques for object detection in satellite and aerial images, followed by conducting experiments in order to implement a genetic algorithm to find the right hyperparameters and then followed by another experiment which compares the performance between optimal and original deep learning model on basis of performance metrics mentioned in the study. Results. The literature review results depict that deep learning techniques for object detection in satellite and aerial images are Faster R-CNN, SSD and YOLO. The results of experiments show that the genetic algorithm was successful in finding optimal hyperparameters. The accuracy achieved by optimized models was higher than the original models in the case of aircraft, vehicles and ship detection. The results also show that the training times for the models have been reduced with the use of optimal hyperparameters with slight decrease in precision when detecting ships. Conclusions. After analyzing all the results carefully, the best deep learning techniques to detect aircraft, vehicles and ships are found and stated. The implementation of the genetic algorithm has been successful as it provided a set of hyperparameters which resulted in the improvement of accuracy, precision and recall in all scenarios except for values of precision in ship detection as well as improvement in training times.
|
146 |
Detekce srdečních buněk v mikroskopickém obrazu / Detection of cardiac cells in microscopic imageMusikhina, Ksenia January 2009 (has links)
This work is devoted to problem of detection of cardiac cells in microscopic picture. All possible means of preprocessing and segmentation were considered with the aim to choose the most suitable method for further classification. Different methods of classification were be testing: method of objects attributes and classifier based on neural network. As a result was obtained the number of living and dead cardiac cells and percentage of them. The electivity of classification methods was calculated by sensitivity and specificity. The user’s interface was created for improvement of clearness classification in MATLAB environment.
|
147 |
Detektor objektů s využitím vlnkové transformace / Wavelet transform based object detectorMikuš, Ondřej January 2009 (has links)
This thesis deals with applying methods on object detection in image. Separation of objects off the background is often needed during the image processing. It isolates the region of interest that can be worked with. The main purpose of this paper is the explanation of principles of pre-processing and segmentation of image, resulting in object detection using the wavelet transformation. This wavelet transformation is described more in detail, because it is the base of the primary used method. In the practical part of this thesis the main method was implemented to MATLAB environment and tested on set of images. The method was tested for robustness against noise and blur of image. It was compared with commonly used methods, using the edge detectors and thresholding. A simulation program was created for comparison of methods efficiency, including user interface.
|
148 |
Lokalizace herních nástrojů / Game instruments localizationČerný, Jakub January 2010 (has links)
This thesis deal with problem of object detection. Object may be of different shapes and colors. The aim of this thesis is to determine the location, direction of movement, angle and proximity.
|
149 |
Sledování objektu ve videosekvenci / Object tracking in videosequenceNešpor, Zdeněk January 2013 (has links)
This thesis deals with tracking a predefined object in the movie. After a brief introduction describes the procedure suitable for the detection of an object in a video sequence, where the methods are also discussed in detail. There is dealt with issues of image preprocessing, image segmentation and object detection in the image. The main emphasis is laid on using detectors of interest points and descriptors of areas - SURF and SIFT. The second part deals with the practical implementation of a program suitable to monitor predefined object in the movie. First are analyzed libraries suitable for object tracking in a video sequence in an environment of Java, followed by a detailed description of the selected library OpenCV along with wrapper JavaCV. Further described is own application in terms of control and functionality are described key method. Outputs along with discussion and evaluation are presented at the end of work.
|
150 |
Boundary Profile Representation for Objects and Their Surroundings in Outdoor VideosCandamo, Joshua 17 August 2009 (has links)
A novel approach to represent the profile of objects using Gaussian models is presented. The profile is a representation of the object and its surrounding regions. The object profile can be viewed as a comprehensive feature of that object and its surrounding regions. Different algorithms to estimate the profile are described. Geometric descriptors of the model are also proposed. The profile model is empirically shown to be effective and easily applicable to certain object recognition and segmentation tasks. Application experiments include modeling thin and thick objects as straight-lines, curves, and blobs using different primitives such as gray-level intensities, RGB, and HSV color. The datasets used for empirical validation are quite challenging. Datasets include images and videos corresponding to outdoor video, most of them with moving cameras. Some of the typical problems faced with the used datasets are: digital scaling, compression artifacts, camera jitter, weather effects, and cluttered backgrounds. We demonstrate the potential of leveraging the context of objects of interest as a part of an online detection process. Sample applications including detection of wires, sea horizon, street, and vehicles in outdoor videos are considered.
|
Page generated in 0.0792 seconds