GCNet: Non-local Networks Meet Squeeze-Excitation Networks and Beyond

25 Apr 2019  ·  Yue Cao, Jiarui Xu, Stephen Lin, Fangyun Wei, Han Hu ·

The Non-Local Network (NLNet) presents a pioneering approach for capturing long-range dependencies, via aggregating query-specific global context to each query position. However, through a rigorous empirical analysis, we have found that the global contexts modeled by non-local network are almost the same for different query positions within an image... In this paper, we take advantage of this finding to create a simplified network based on a query-independent formulation, which maintains the accuracy of NLNet but with significantly less computation. We further observe that this simplified design shares similar structure with Squeeze-Excitation Network (SENet). Hence we unify them into a three-step general framework for global context modeling. Within the general framework, we design a better instantiation, called the global context (GC) block, which is lightweight and can effectively model the global context. The lightweight property allows us to apply it for multiple layers in a backbone network to construct a global context network (GCNet), which generally outperforms both simplified NLNet and SENet on major benchmarks for various recognition tasks. The code and configurations are released at https://github.com/xvjiarui/GCNet. read more

PDF Abstract

Datasets


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Object Detection COCO minival GCNet (ResNeXt-101 + DCN + cascade + GC r16) box AP 47.9 # 17
AP50 66.9 # 14
AP75 52.2 # 12
Instance Segmentation COCO minival GCNet (ResNeXt-101 + DCN + cascade + GC r16) mask AP 40.9 # 18
Object Detection COCO minival GCnet (ResNet-50-FPN, GRoIE) box AP 40.3 # 77
AP50 62.4 # 34
AP75 44 # 47
APS 24.2 # 41
APM 44.4 # 40
APL 52.5 # 50
Object Detection COCO test-dev GCNet (ResNeXt-101 + DCN + cascade + GC r4) box AP 48.4 # 47
AP50 67.6 # 42
AP75 52.7 # 42
Instance Segmentation COCO test-dev GCNet (ResNeXt-101 + DCN + cascade + GC r16) mask AP 41.5% # 14

Methods