Local-Global Context Aware Transformer for Language-Guided Video Segmentation

18 Mar 2022  ·  Chen Liang, Wenguan Wang, Tianfei Zhou, Jiaxu Miao, Yawei Luo, Yi Yang ·

We explore the task of language-guided video segmentation (LVS). Previous algorithms mostly adopt 3D CNNs to learn video representation, struggling to capture long-term context and easily suffering from visual-linguistic misalignment. In light of this, we present Locater (local-global context aware Transformer), which augments the Transformer architecture with a finite memory so as to query the entire video with the language expression in an efficient manner. The memory is designed to involve two components -- one for persistently preserving global video content, and one for dynamically gathering local temporal context and segmentation history. Based on the memorized local-global context and the particular content of each frame, Locater holistically and flexibly comprehends the expression as an adaptive query vector for each frame. The vector is used to query the corresponding frame for mask generation. The memory also allows Locater to process videos with linear time complexity and constant size memory, while Transformer-style self-attention computation scales quadratically with sequence length. To thoroughly examine the visual grounding capability of LVS models, we contribute a new LVS dataset, A2D-S+, which is built upon A2D-S dataset but poses increased challenges in disambiguating among similar objects. Experiments on three LVS datasets and our A2D-S+ show that Locater outperforms previous state-of-the-arts. Further, we won the 1st place in the Referring Video Object Segmentation Track of the 3rd Large-scale Video Object Segmentation Challenge, where Locater served as the foundation for the winning solution. Our code and dataset are available at: https://github.com/leonnnop/Locater

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Referring Expression Segmentation A2D Sentences Locater Precision@0.5 0.709 # 8
Precision@0.9 0.101 # 12
IoU overall 0.69 # 9
IoU mean 0.597 # 11
Precision@0.6 0.64 # 10
Precision@0.7 0.525 # 10
Precision@0.8 0.351 # 11
AP 0.465 # 7
Referring Expression Segmentation Refer-YouTube-VOS (2021 public validation) Locater J&F 50 # 22
J 48.8 # 22
F 51.1 # 21