Abstract
Modern neural network algorithms for object detection tasks require a large la-belled dataset for training. In a number of practical applications creation and an-notation of large data, collections requires considerable resources which are not always available. One of the solutions to this problem is creation of artificial images containing the object of interest. In this work the use of generative adversarial networks (GAN) for generation of images of target objects is proposed. It is demonstrated experimentally that GAN’s allows to create new images on the basis of the initial collection of real images on background images (not containing objects), which simulate real images accurately enough. Due to this, it is possible to create a new training collection containing a greater variety of training examples, which allows to achieve higher precision for detection algorithm. In our setting, GAN training does not require more data than is required for direct detector training. The proposed method has been tested to teach a network for detecting unmanned aerial vehicles (UAVs).
Highlights
The majority of modern object detection systems and computer vision algorithms are based on machine learning, primarily neural networks
In order to achieve high accuracy of detector training in conditions of very limited training set, we have investigated the possibility of using generative adversarial neural networks (GAN) for creation of synthetic training images
The second experiment repeated the method proposed in the article [1]. The data for this experiment were extended by placing drone renders in arbitrary places
Summary
The majority of modern object detection systems and computer vision algorithms are based on machine learning, primarily neural networks. They have proven their reliability and quality in a wide range of tasks. The main disadvantage of such algorithms is the requirement of large (or even super large) annotated training datasets. In case of training the detector for a specific object that is not represented in large public annotated data collections, or the need to work in specific conditions. In the absence of data from satellite navigation systems and requirements for the absence of additional radio wave sources, the use of passive sensors such as video cameras with detection algorithm is extremely useful
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callMore From: Proceedings of the 30th International Conference on Computer Graphics and Machine Vision (GraphiCon 2020). Part 2
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
