End-to-End Video Object Detection with Spatial-Temporal Transformers скачать в хорошем качестве

End-to-End Video Object Detection with Spatial-Temporal Transformers

End-to-End Video Object Detection with Spatial-Temporal Transformers (Machine Learning Research Paper Explained) #machinelearning #transformers #objectdetection #pointclouds Sources: https://arxiv.org/abs/2105.10920 https://arxiv.org/abs/2005.12872 https://arxiv.org/abs/2010.04159 DETR has been recently proposed to eliminate the need for many hand-designed components in object detection while demonstrating good performance. However, it suffers from slow convergence and limited feature spatial resolution, due to the limitation of Transformer attention modules in processing image feature maps. To mitigate these issues, we proposed Deformable DETR, whose attention modules only attend to a small set of key sampling points around a reference. Deformable DETR can achieve better performance than DETR (especially on small objects) with 10 times less training epochs. Extensive experiments on the COCO benchmark demonstrate the effectiveness of our approach. Code is released at this https URL. 0:00 Intro 0:28 Problem formulation 3:08 Overview 3:36 Architecture overview 6:06 Architecture deep-dive 32:53 Architecture Review 34:47 Results 41:33 Outro