New AI model Bar-JEPA extracts values from bar charts using self-supervised learning

Researchers have developed Bar-JEPA, a novel approach for computationally extracting numerical values from bar charts. This method utilizes a Joint-Embedding Predictive Architecture (JEPA) encoder to generate semantically rich latent features in a self-supervised manner, addressing the scarcity of labeled data for chart de-rendering. A simple decoder trained on these features can then recover bar values by outputting tick and bar coordinates, demonstrating effectiveness compared to traditional supervised baselines. The project's code, datasets, and checkpoints are publicly available on GitHub. AI

IMPACT This method could improve automated data extraction from visual sources, aiding in data analysis and accessibility.

RANK_REASON Academic paper detailing a new model and methodology. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• Bar-JEPA
• GitHub
• Hugging Face

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →