Seq-NMS for Video Object Detection
This addresses video object detection challenges for computer vision applications, but it is incremental as it builds on existing single image detection methods.
The paper tackled the problem of inconsistent object detection across frames in videos by modifying the post-processing phase to use high-scoring detections from nearby frames to boost weaker ones, achieving superior results to state-of-the-art single image techniques and placing 3rd in the ILSVRC2015 video object detection task.
Video object detection is challenging because objects that are easily detected in one frame may be difficult to detect in another frame within the same clip. Recently, there have been major advances for doing object detection in a single image. These methods typically contain three phases: (i) object proposal generation (ii) object classification and (iii) post-processing. We propose a modification of the post-processing phase that uses high-scoring object detections from nearby frames to boost scores of weaker detections within the same clip. We show that our method obtains superior results to state-of-the-art single image object detection techniques. Our method placed 3rd in the video object detection (VID) task of the ImageNet Large Scale Visual Recognition Challenge 2015 (ILSVRC2015).