Dynamic R-CNN: Towards High Quality Object Detection via Dynamic Training

Although two-stage object detectors have continuously advanced the state-of-the-art performance in recent years, the training process itself is far from crystal. In this work, we first point out the inconsistency problem between the fixed network settings and the dynamic training procedure, which greatly affects the performance. For example, the fixed label assignment strategy and regression loss function cannot fit the distribution change of proposals and thus are harmful to training high quality detectors. Consequently, we propose Dynamic R-CNN to adjust the label assignment criteria (IoU threshold) and the shape of regression loss function (parameters of SmoothL1 Loss) automatically based on the statistics of proposals during training. This dynamic design makes better use of the training samples and pushes the detector to fit more high quality samples. Specifically, our method improves upon ResNet-50-FPN baseline with 1.9% AP and 5.5% AP$_{90}$ on the MS COCO dataset with no extra overhead. Codes and models are available at https://github.com/hkzhang95/DynamicRCNN.

PDF Abstract ECCV 2020 PDF ECCV 2020 Abstract

Datasets


Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Object Detection COCO test-dev Dynamic R-CNN (ResNet-101-DCN, multi-scale) box mAP 50.1 # 76
AP50 68.3 # 43
AP75 55.6 # 30
APS 32.8 # 25
APM 53.0 # 30
APL 61.2 # 41
Hardware Burden None # 1
Operations per network pass None # 1

Methods