Summary of Learning Disentangled Representation in Object-centric Models For Visual Dynamics Prediction Via Transformers, by Sanket Gandhi et al.
Learning Disentangled Representation in Object-Centric Models for Visual Dynamics Prediction via Transformers
by Sanket Gandhi, Atul, Samanyu Mahajan, Vishal Sharma, Rushil Gupta, Arnab Kumar Mondal, Parag Singla
First submitted to arxiv on: 3 Jul 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 A recent study has shown that using object-centric representations can greatly improve the accuracy and interpretability of learning dynamics models. This work takes this idea further by asking whether learning a disentangled representation can improve the accuracy of predicting visual dynamics in object-centric models. The authors create an architecture that represents objects as linear combinations of learnable concept vectors, which are refined during the learning process. They use self-attention to predict the next state, discovering semantically meaningful blocks and improving the accuracy of dynamics prediction compared to state-of-the-art (SOTA) object-centric models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Learning a disentangled representation can greatly help improve the accuracy of predicting visual dynamics in object-centric models. This is done by creating an architecture that represents objects as linear combinations of learnable concept vectors, which are refined during the learning process. The authors use self-attention to predict the next state and discover semantically meaningful blocks. |
Keywords
» Artificial intelligence » Self attention
