Summary of Two Heads Are Better Than One: Neural Networks Quantization with 2d Hilbert Curve-based Output Representation, by Mykhailo Uss et al.
Two Heads are Better Than One: Neural Networks Quantization with 2D Hilbert Curve-based Output Representation
by Mykhailo Uss, Ruslan Yermolenko, Olena Kolodiazhna, Oleksii Shashko, Ivan Safonov, Volodymyr Savin, Yoonjae Yeo, Seowon Ji, Jaeyun Jeong
First submitted to arxiv on: 22 May 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 approach for deep neural networks (DNN) quantization uses a redundant representation of DNN’s output. This is achieved by representing the target quantity as a point on a 2D parametric curve. The modified DNN model predicts 2D points that are mapped back to the target quantity at a post-processing stage, resulting in reduced quantization error. The technique demonstrates an approximately 5-fold reduction in quantization error for the INT8 model at both CPU and DSP delegates, with minimal inference time increase (less than 7%). This approach has potential applications in various tasks such as segmentation, object detection, key-points prediction, Depth-From-Stereo, and more. The method utilizes U-Net architecture and vision transformer models, showcasing its versatility. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper introduces a new way to make deep neural networks (DNNs) work better on devices with limited memory or power. They do this by changing how the DNN predicts the output. Instead of predicting just one number, it predicts two numbers that are used to find the correct answer later. This helps reduce mistakes made when converting the DNN from a big number system to a smaller one (quantization). The method works well for certain types of tasks, such as looking at depth in images or detecting objects. |
Keywords
» Artificial intelligence » Inference » Object detection » Quantization » Vision transformer
