Summary of Sltnet: Efficient Event-based Semantic Segmentation with Spike-driven Lightweight Transformer-based Networks, by Xiaxin Zhu et al.
SLTNet: Efficient Event-based Semantic Segmentation with Spike-driven Lightweight Transformer-based Networks
by Xiaxin Zhu, Fangming Guo, Xianlei Long, Qingyi Gu, Chao Chen, Fuqiang Gu
First submitted to arxiv on: 17 Dec 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The authors introduce SLTNet, a lightweight transformer-based network for event-based semantic segmentation in autonomous driving and robotics. Current ANN-based methods are limited by high computational demands, image frame requirements, and energy consumption. SLTNet addresses these issues with spike-driven convolution blocks (SCBs) for efficient feature extraction and novel spike-driven transformer blocks (STBs) for long-range contextural feature interaction. The model uses a single-branch architecture, maintaining low energy consumption while outperforming state-of-the-art SNN-based methods by up to 9.06% and 9.39% mIoU on DDD17 and DSEC-Semantic datasets respectively. SLTNet also achieves extremely low energy consumption (4.58x) and high inference speed (114 FPS). |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary SLTNet is a new way to help computers see and understand the world around them, especially in situations where there are lots of things moving quickly. Right now, most computer vision models need a lot of power and time to work, which makes them hard to use in places like cars or robots. SLTNet is different because it can do the same job using much less energy and time. This makes it useful for applications that require low power consumption, such as autonomous driving or robotics. |
Keywords
» Artificial intelligence » Feature extraction » Inference » Semantic segmentation » Transformer
