FWL-ToPM Enables Real-Time Full-Waveform LiDAR Transformers
Original title: Towards Real-Time Full-Waveform LiDAR Transformers via Intensity-Guided Token Reduction and Physics-Aware Augmentation
FWL-ToPM provides over 8x speedup and nearly 9-point F1 gain on Ghost-FWL benchmark.
Full-waveform LiDAR captures detailed return waveforms but incurs high computational costs from redundant tokens. FWL-ToPM uses intensity-guided token pruning and merging combined with FWLAug data augmentation. These techniques achieve real-time throughput with accuracy gains and better generalization to unseen scenes on Ghost-FWL.
- Over 8 times speedup at real-time throughput
- Nearly 9-point F1 gain over prior state of the art
- Improves performance on out-of-distribution test set
- Code and OOD dataset publicly available
- Organizations
- Keio University
- AI models
- FWL-ToPM
From the original article
1 min read at huggingface.co
LiDAR is a critical sensor for autonomous driving and robotics, yet it remains vulnerable to adverse conditions such as fog, rain, transparent objects, and low-reflectance surfaces. Full-waveform LiDAR (FWL) addresses these limitations by capturing the complete return waveform as a histogram, preserving rich reflection characteristics including multi-path echoes and weak signals.
By Kotaro Oishi, Kazuma Ikeda, Ryosei Hara · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.