Object Detection in the Image Recognition Process Using Transformers

Modern object detection methods in the image recognition process us-ing transformer technology are analyzed.The various advantages and disadvantages identified. An own network was created based on DETR from FAIR team, its operation analyzed. A comparison of networks perfor-mance with optimized architectures convolutional neural is made.The cloud computing tools, graphics processors, Internet Things clusters or embedded microprocessor systems were used research process.To ensure high detector accuracy real-time results different types devices, an efficient model scaling technique required.The learning illustrated step-by-step process.