Multiframe CenterNet Heatmap ROI Aggregation for Real-Time Video Object Detection

Though Two-stage video object detectors cannot perform detection in real time, the accuracy of them is normally higher than that one-stage detectors. One essence two-stage can easily use feature information from adjacent frames to augment key frame features. How extract and exploit temporal features stream for needs further exploration. CenterNet an anchor-free detector regress bounding boxes heatmap peaks. We propose detected peaks regressed which encompass peak points determine ROIs as extracted A new relation module designed evaluate similarity ROI output effectively In sequence multiple are aggregated a frame. Compared other CenterNet-based detectors, our method achieves improved online real-time performance on ImageNet VID dataset with 78.8% mAP at 36 FPS.