Feature Selective Anchor-Free Module for Single-Shot Object Detection

CVPR 2019  ·  Chenchen Zhu, Yihui He, Marios Savvides ·

We motivate and present feature selective anchor-free (FSAF) module, a simple and effective building block for single-shot object detectors. It can be plugged into single-shot detectors with feature pyramid structure. The FSAF module addresses two limitations brought up by the conventional anchor-based detection: 1) heuristic-guided feature selection; 2) overlap-based anchor sampling. The general concept of the FSAF module is online feature selection applied to the training of multi-level anchor-free branches. Specifically, an anchor-free branch is attached to each level of the feature pyramid, allowing box encoding and decoding in the anchor-free manner at an arbitrary level. During training, we dynamically assign each instance to the most suitable feature level. At the time of inference, the FSAF module can work jointly with anchor-based branches by outputting predictions in parallel. We instantiate this concept with simple implementations of anchor-free branches and online feature selection strategy. Experimental results on the COCO detection track show that our FSAF module performs better than anchor-based counterparts while being faster. When working jointly with anchor-based branches, the FSAF module robustly improves the baseline RetinaNet by a large margin under various settings, while introducing nearly free inference overhead. And the resulting best model can achieve a state-of-the-art 44.6% mAP, outperforming all existing single-shot detectors on COCO.

PDF Abstract CVPR 2019 PDF CVPR 2019 Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Object Detection COCO minival FSAF (ResNeXt-101, anchor-based branches) box AP 41.6 # 161
AP50 62.4 # 70
Object Detection COCO minival FSAF (ResNet-50) box AP 35.9 # 204
AP50 55.0 # 113
AP75 37.9 # 105
APS 19.8 # 87
APM 39.6 # 88
APL 48.2 # 93
Object Detection COCO minival FSAF (ResNet-101) box AP 37.9 # 199
AP50 58.0 # 105
Object Detection COCO minival FSAF (ResNet-101, anchor-based branches) box AP 39.3 # 187
AP50 59.2 # 97
Object Detection COCO test-dev FSAF (ResNeXt-101, multi-scale) box mAP 44.6 # 139
AP50 65.2 # 76
AP75 48.6 # 84
APS 29.7 # 53
APM 47.1 # 87
APL 54.6 # 105
Hardware Burden None # 1
Operations per network pass None # 1
Object Detection COCO test-dev FSAF (ResNet-101, single-scale) box mAP 40.9 # 182
AP50 61.5 # 120
AP75 44 # 133
APS 24 # 105
APM 44.2 # 116
APL 51.3 # 128
Hardware Burden 38G # 1
Operations per network pass None # 1

Methods