Summary of Fora: Low-rank Adaptation Model Beyond Multimodal Siamese Network, by Weiying Xie and Yusi Zhang and Tianlin Hui and Jiaqing Zhang and Jie Lei and Yunsong Li
FoRA: Low-Rank Adaptation Model beyond Multimodal Siamese Network
by Weiying Xie, Yusi Zhang, Tianlin Hui, Jiaqing Zhang, Jie Lei, Yunsong Li
First submitted to arxiv on: 23 Jul 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 Low-rank Modal Adaptors (LMA) multimodal object detector addresses the challenges of existing two-stream backbone networks by leveraging a shared backbone and lightweight modal adaptors. This approach enhances consistency through shared parameters, while modality-specific features are learned through adaptive rank allocation. The method is evaluated on two datasets, achieving a 10.4% accuracy improvement over the state-of-the-art method with a reduction in model size (from 159M to 149M parameters). The proposed LMA outperforms existing methods, particularly on the DroneVehicle dataset. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Multimodal object detection is important for detecting objects in different conditions. This paper proposes a new way to do this called Low-rank Modal Adaptors (LMA). It uses a shared part of the network and separate parts for different types of data. This helps learn features that are specific to each type of data, while also sharing information between them. The method is tested on two datasets and shows improvement over previous methods. |
Keywords
* Artificial intelligence * Object detection
