Summary of Altermoma: Fusion Redundancy Pruning For Camera-lidar Fusion Models with Alternative Modality Masking, by Shiqi Sun et al.
AlterMOMA: Fusion Redundancy Pruning for Camera-LiDAR Fusion Models with Alternative Modality Masking
by Shiqi Sun, Yantao Lu, Ning Liu, Bo Jiang, JinChao Chen, Ying Zhang
First submitted to arxiv on: 26 Sep 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 proposed fusion mechanism in camera-LiDAR models significantly enhances perception performance in autonomous driving by leveraging the strengths of each modality while minimizing their weaknesses. Additionally, pre-trained backbones are utilized for efficient training. However, directly loading single-modal pre-trained camera and LiDAR backbones into camera-LiDAR fusion models introduces similar feature redundancy across modalities due to the nature of the fusion mechanism. A novel pruning framework called Alternative Modality Masking Pruning (AlterMOMA) is proposed to address this issue. AlterMOMA employs alternative masking on each modality and identifies redundant parameters by observing the reactivation process when one modality’s parameters are masked. The redundant parameters can be pruned using an importance score evaluation function, Alternative Evaluation (AlterEva), which evaluates loss changes when certain modality parameters are activated or deactivated. Experimental results on nuScene and KITTI datasets demonstrate that AlterMOMA outperforms existing pruning methods, achieving state-of-the-art performance. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary In simple terms, this paper is about improving the performance of camera-LiDAR fusion models used in self-driving cars. The researchers found a problem where using pre-trained models from single-modal cameras and LiDAR sensors introduced unnecessary information that made it harder to work well together. To fix this, they developed a new way to identify and remove redundant information called Alternative Modality Masking Pruning (AlterMOMA). This method works by temporarily disabling one modality’s features and seeing if the other modality’s features can still perform well. By doing so, AlterMOMA can efficiently prune away unnecessary information, leading to better performance. |
Keywords
» Artificial intelligence » Pruning
