Temporally Distributed Networks for Fast Video Semantic Segmentation
This work addresses the need for efficient video segmentation in applications like autonomous driving, offering a novel method that improves speed without sacrificing accuracy.
The paper tackles the problem of fast video semantic segmentation by proposing TDNet, which distributes sub-networks over sequential frames to reduce computation, achieving state-of-the-art accuracy with significantly faster speed and lower latency on datasets like Cityscapes, CamVid, and NYUD-v2.
We present TDNet, a temporally distributed network designed for fast and accurate video semantic segmentation. We observe that features extracted from a certain high-level layer of a deep CNN can be approximated by composing features extracted from several shallower sub-networks. Leveraging the inherent temporal continuity in videos, we distribute these sub-networks over sequential frames. Therefore, at each time step, we only need to perform a lightweight computation to extract a sub-features group from a single sub-network. The full features used for segmentation are then recomposed by application of a novel attention propagation module that compensates for geometry deformation between frames. A grouped knowledge distillation loss is also introduced to further improve the representation power at both full and sub-feature levels. Experiments on Cityscapes, CamVid, and NYUD-v2 demonstrate that our method achieves state-of-the-art accuracy with significantly faster speed and lower latency.