Summary of Instance-aware Group Quantization For Vision Transformers, by Jaehyeon Moon et al.
Instance-Aware Group Quantization for Vision Transformers
by Jaehyeon Moon, Dohyung Kim, Junyong Cheon, Bumsub Ham
First submitted to arxiv on: 1 Apr 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 In this paper, researchers introduce a novel model compression technique called instance-aware group quantization for vision transformers (IGQ-ViT). The goal is to efficiently compress pre-trained vision transformers while maintaining their performance. The authors identify that directly applying post-training quantization (PTQ) methods designed for convolutional neural networks (CNNs) to vision transformers incurs significant performance degradation. To address this, the researchers propose a new method that dynamically groups activation maps into multiple instances based on statistical properties, allowing for efficient compression of vision transformer models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This technique can be used for various applications such as image classification, object detection, and instance segmentation. The authors show extensive experimental results demonstrating the effectiveness of their approach. |
Keywords
* Artificial intelligence * Image classification * Instance segmentation * Model compression * Object detection * Quantization * Vision transformer * Vit
