Benchmarking Vision-Language Models on Optical Character Recognition in Dynamic Video Environments · HackerPapers discussions

Benchmarking Vision-Language Models on Optical Character Recognition in Dynamic Video Environments — HN submission

Fri, 14 Feb 2025 07:26:16 GMT

Benchmarking vision-language models on OCR in dynamic video environments

142 points; 58 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Benchmarking Vision-Language Models on Optical Character Recognition in Dynamic Video Environments — mentioned by ashu_trv

Thu, 13 Feb 2025 12:44:06 GMT

A new benchmark study evaluates Vision-Language Models (Claude-3, Gemini-1.5, GPT-4o) against traditional OCR tools (EasyOCR, RapidOCR) for extracting text from videos. The findings show VLMs outperforming OCR in many cases but also highlight challenges like hallucinated text and handling occluded/stylized fonts. The dataset (1,477 manually annotated frames) and benchmarking framework are publicly available to encourage further research. Paper: https://arxiv.org/abs/2502.06445 Dataset & Repo: https://github.com/video-db/ocr-benchmark Would love to hear thoughts from the community on the…

Paper and discussion history. A mention is not an endorsement.

Benchmarking Vision-Language Models on Optical Character Recognition in Dynamic Video Environments — HN submission

Thu, 13 Feb 2025 12:44:06 GMT

Vision-Language Models vs. Traditional OCR in Video – New Benchmark

6 points; 1 discussion reply on HN. Paper and source history. A submission is not an endorsement.