Post
1116
π Introducing ColFlor: An Efficient, OCR-Free Vision-Language Document Retrieval Model π
Earlier this year, ColPali revolutionized document retrieval by eliminating the need for error-prone OCR pipelines. Instead, it directly processes the document images. However, with its 3 billion parameters, ColPali is computationally heavy for large-scale applications.
Thatβs where ColFlor comes inβa smaller, faster alternative! π At 17x smaller than ColPali, ColFlor offers a more efficient, OCR-free document retrieval solution, making it ideal for users with limited computing resources (GPU Poor). π‘
Key Highlights:
π§ 174M parameters (vs. 3B for ColPali)
β‘ 9.8x faster query encoding, 5.25x faster image encoding
π Only 1.8% performance drop on text-rich English documents
Check out the full blog post for more insights on modeling, training, and evaluations across various document retrieval tasks! π
Also, feel free to try our demo on huggingface π€
π Resources:
π Blog post: https://huggingface.co/blog/ahmed-masry/colflor
π§ Model: ahmed-masry/ColFlor
π» Demo: ahmed-masry/ColFlor-Demo
ποΈββοΈ Training code: https://github.com/AhmedMasryKU/colflor
π Evaluation code: https://github.com/AhmedMasryKU/vidore-benchmark-colflor
Earlier this year, ColPali revolutionized document retrieval by eliminating the need for error-prone OCR pipelines. Instead, it directly processes the document images. However, with its 3 billion parameters, ColPali is computationally heavy for large-scale applications.
Thatβs where ColFlor comes inβa smaller, faster alternative! π At 17x smaller than ColPali, ColFlor offers a more efficient, OCR-free document retrieval solution, making it ideal for users with limited computing resources (GPU Poor). π‘
Key Highlights:
π§ 174M parameters (vs. 3B for ColPali)
β‘ 9.8x faster query encoding, 5.25x faster image encoding
π Only 1.8% performance drop on text-rich English documents
Check out the full blog post for more insights on modeling, training, and evaluations across various document retrieval tasks! π
Also, feel free to try our demo on huggingface π€
π Resources:
π Blog post: https://huggingface.co/blog/ahmed-masry/colflor
π§ Model: ahmed-masry/ColFlor
π» Demo: ahmed-masry/ColFlor-Demo
ποΈββοΈ Training code: https://github.com/AhmedMasryKU/colflor
π Evaluation code: https://github.com/AhmedMasryKU/vidore-benchmark-colflor