Summary of Unsupervised Domain Adaptation Approaches For Chessboard Recognition, by Wassim Jabbour et al.
Unsupervised Domain Adaptation Approaches for Chessboard Recognition
by Wassim Jabbour, Enzo Benoit-Jeannin, Oscar Bedford, Saif Shahin
First submitted to arxiv on: 19 Oct 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 The proposed end-to-end pipeline employs domain adaptation to predict labels of real, top-view, unlabeled chessboard images using synthetic, labeled images. The pipeline consists of a pre-processing phase detecting the board, cropping individual squares, and feeding them one at a time to a Deep Learning model. The model predicts square labels, which are then passed to a post-processing pipeline generating Forsyth-Edwards Notation (FEN) of the position. Three approaches were considered: fine-tuning VGG16 on ImageNet; fine-tuning with CORAL loss; and Domain Adversarial Neural Network (DANN). The DANN model achieved 97% accuracy, only losing 3% compared to the baseline trained directly on the target domain. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps automate record-keeping in chess by using computer vision. Right now, players need to manually take photos of their board and label each square, which takes up a lot of time. To solve this problem, the authors created an end-to-end system that uses fake labeled images to teach a computer model what different squares look like. The model can then predict what’s on each square in real-time. Three different approaches were tried: one was just fine-tuned from pre-trained model, another used a special loss function to help it adapt, and the third used a special neural network that helps it learn from both fake and real images. The best approach lost only 3% accuracy compared to using all the labeled data. |
Keywords
» Artificial intelligence » Deep learning » Domain adaptation » Fine tuning » Loss function » Neural network
