About
PathGen-1.6M is a large-scale pathology image-text dataset generated through multi-agent collaboration. The framework employs multiple specialized agents to create high-quality image-text pairs from pathology slides, resulting in 1.6 million pairs that can be used for training vision-language models in digital pathology. Published at ICLR 2025 as an Oral presentation.
Tech Stack
PythonPyTorchMulti-Agent SystemVision-Language Models
Research Paper
View PaperQuick Start
git clone https://github.com/PathFoundation/PathGen-1.6M.git && cd PathGen-1.6M && pip install -r requirements.txt