Reliable Propagation-Correction Modulation for Video Object Segmentation

6 Dec 2021  ·  Xiaohao Xu, Jinglu Wang, Xiao Li, Yan Lu ·

Error propagation is a general but crucial problem in online semi-supervised video object segmentation. We aim to suppress error propagation through a correction mechanism with high reliability. The key insight is to disentangle the correction from the conventional mask propagation process with reliable cues. We introduce two modulators, propagation and correction modulators, to separately perform channel-wise re-calibration on the target frame embeddings according to local temporal correlations and reliable references respectively. Specifically, we assemble the modulators with a cascaded propagation-correction scheme. This avoids overriding the effects of the reliable correction modulator by the propagation modulator. Although the reference frame with the ground truth label provides reliable cues, it could be very different from the target frame and introduce uncertain or incomplete correlations. We augment the reference cues by supplementing reliable feature patches to a maintained pool, thus offering more comprehensive and expressive object representations to the modulators. In addition, a reliability filter is designed to retrieve reliable patches and pass them in subsequent frames. Our model achieves state-of-the-art performance on YouTube-VOS18/19 and DAVIS17-Val/Test benchmarks. Extensive experiments demonstrate that the correction mechanism provides considerable performance gain by fully utilizing reliable guidance. Code is available at: https://github.com/JerryX1110/RPCMVOS.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Semi-Supervised Video Object Segmentation DAVIS 2016 RPCMVOS Jaccard (Mean) 87.1 # 41
F-measure (Mean) 94 # 8
J&F 90.6 # 26
Semi-Supervised Video Object Segmentation DAVIS 2017 (test-dev) RPCMVOS J&F 79.2 # 22
Jaccard (Mean) 75.8 # 22
F-measure (Mean) 82.6 # 22
Video Object Segmentation DAVIS 2017 (test-dev) RPCMVOS Jaccard 75.8 # 3
F-measure 82.6 # 3
Mean Jaccard & F-Measure 79.2 # 3
Semi-Supervised Video Object Segmentation DAVIS 2017 (test-dev) RPCMVOS-Full-Res J&F 81 # 15
Jaccard (Mean) 77.6 # 13
F-measure (Mean) 84.3 # 17
Semi-Supervised Video Object Segmentation DAVIS 2017 (val) RPCMVOS Jaccard (Mean) 81.3 # 28
F-measure (Mean) 86 # 32
J&F 83.7 # 30
Semi-Supervised Video Object Segmentation YouTube-VOS 2018 RPCMVOS-MS F-Measure (Seen) 87.9 # 23
F-Measure (Unseen) 86.9 # 19
Overall 84.3 # 21
Jaccard (Seen) 83.3 # 21
Jaccard (Unseen) 78.9 # 14
Semi-Supervised Video Object Segmentation YouTube-VOS 2018 RPCMVOS F-Measure (Seen) 87.7 # 27
F-Measure (Unseen) 86.7 # 21
Overall 84 # 24
Jaccard (Seen) 83.1 # 24
Speed (FPS) 78.5 # 1
Semi-Supervised Video Object Segmentation YouTube-VOS 2019 RPCMVOS Overall 83.9 # 17
Jaccard (Seen) 82.6 # 17
Jaccard (Unseen) 79.1 # 14
F-Measure (Seen) 86.9 # 18
F-Measure (Unseen) 87.1 # 15
Video Object Segmentation YouTube-VOS 2019 RPCMVOS Mean Jaccard & F-Measure 83.9 # 5
Jaccard (Seen) 82.6 # 6
Jaccard (Unseen) 79.1 # 3
F-Measure (Seen) 86.9 # 6
F-Measure (Unseen) 87.1 # 4

Methods


No methods listed for this paper. Add relevant methods here