Summary of Keypoint Aware Masked Image Modelling, by Madhava Krishna et al.
Keypoint Aware Masked Image Modelling
by Madhava Krishna, A V Subramanyam
First submitted to arxiv on: 18 Jul 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
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 SimMIM, a widely used method for pretraining vision transformers, has been shown to perform sub-optimally in linear probing. To address this, we propose an efficient patch-wise weighting derived from keypoint features that captures local information and provides better context during SimMIM’s reconstruction phase. Our method, KAMIM, improves top-1 linear probing accuracy from 16.12% to 33.97%, and finetuning accuracy from 76.78% to 77.3%, when tested on the ImageNet-1K dataset with a ViT-B model. We conduct extensive testing on different datasets, keypoint extractors, and model architectures, observing that patch-wise weighting augments linear probing performance for larger pretraining datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary KAMIM is a new way to improve SimMIM’s performance in linear probing tasks. It works by using special weights for each patch of an image, based on the locations of key points in that image. This helps the model understand more about what it’s seeing and make better predictions. We tested KAMIM on several different datasets and models, and it worked well across the board. The results show that KAMIM can help improve performance by a lot, especially when using large datasets. |
Keywords
* Artificial intelligence * Pretraining * Vit
