BoxInst: High-Performance Instance Segmentation with Box Annotations

CVPR 2021  ·  Zhi Tian, Chunhua Shen, Xinlong Wang, Hao Chen ·

We present a high-performance method that can achieve mask-level instance segmentation with only bounding-box annotations for training. While this setting has been studied in the literature, here we show significantly stronger performance with a simple design (e.g., dramatically improving previous best reported mask AP of 21.1% in Hsu et al. (2019) to 31.6% on the COCO dataset). Our core idea is to redesign the loss of learning masks in instance segmentation, with no modification to the segmentation network itself. The new loss functions can supervise the mask training without relying on mask annotations. This is made possible with two loss terms, namely, 1) a surrogate term that minimizes the discrepancy between the projections of the ground-truth box and the predicted mask; 2) a pairwise loss that can exploit the prior that proximal pixels with similar colors are very likely to have the same category label. Experiments demonstrate that the redesigned mask loss can yield surprisingly high-quality instance masks with only box annotations. For example, without using any mask annotations, with a ResNet-101 backbone and 3x training schedule, we achieve 33.2% mask AP on COCO test-dev split (vs. 39.1% of the fully supervised counterpart). Our excellent experiment results on COCO and Pascal VOC indicate that our method dramatically narrows the performance gap between weakly and fully supervised instance segmentation. Code is available at: https://git.io/AdelaiDet

PDF Abstract CVPR 2021 PDF CVPR 2021 Abstract

Datasets


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Box-supervised Instance Segmentation COCO test-dev BoxInst mask AP 35.0 # 4
Weakly-supervised instance segmentation COCO test-dev BoxInst (ResNet-101-DCN-BiFPN) AP 35.0 # 3
AP@50 59.3 # 3
AP@75 35.6 # 3
AP@S 17.1 # 2
AP@M 37.2 # 3
AP@L 48.9 # 3
Weakly-supervised instance segmentation COCO test-dev BoxInst (ResNet-101-BiFPN) AP 33.9 # 4
AP@50 57.7 # 4
AP@75 34.5 # 4
AP@S 16.5 # 4
AP@M 36.1 # 4
AP@L 46.6 # 5
Weakly-supervised instance segmentation COCO test-dev BoxInst (ResNet-101-FPN) AP 33.2 # 5
AP@50 56.5 # 5
AP@75 33.6 # 5
AP@S 16.2 # 5
AP@M 35.3 # 5
AP@L 45.1 # 6
Weakly-supervised instance segmentation COCO test-dev BoxInst (ResNet-50-FPN) AP 32.1 # 6
AP@50 55.1 # 6
AP@75 32.4 # 7
AP@S 15.6 # 6
AP@M 34.3 # 6
AP@L 43.5 # 7
Box-supervised Instance Segmentation PASCAL VOC 2012 val BoxInst mask AP 36.5 # 2
AP_50 61.4 # 4
AP_75 37.0 # 3

Methods


No methods listed for this paper. Add relevant methods here