Summary of Integrated Image-text Based on Semi-supervised Learning For Small Sample Instance Segmentation, by Ruting Chi et al.
Integrated Image-Text Based on Semi-supervised Learning for Small Sample Instance Segmentation
by Ruting Chi, Zhiyi Huang, Yuexing Han
First submitted to arxiv on: 21 Oct 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 novel solution for small sample instance segmentation addresses the limitations of existing methods that rely on meta-learning by pre-training models on support sets and fine-tuning on query sets. The approach maximizes the utilization of existing information without increasing annotation burden or training costs. It consists of two modules: generating pseudo labels to increase available samples and integrating text and image features for accurate classification. This method is suitable for both box-free and box-dependent frameworks, reducing reliance on pre-training. Experiments conducted on three datasets from different scenes (on land, underwater, and under microscope) demonstrate the effectiveness and superiority of the proposed method in improving small sample instance segmentation performance. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A new way to segment objects in small groups has been developed. This approach is better than existing methods that need extra training time and special data sets. The solution uses two tools: one generates fake labels for unlabeled data, increasing the number of samples available, and the other combines image and text information to improve classification accuracy. This method works with both types of frameworks (box-free or box-dependent) and doesn’t require as much pre-training. Tests on three different datasets show that this approach is effective and better than others. |
Keywords
» Artificial intelligence » Classification » Fine tuning » Instance segmentation » Meta learning
