search

Q-Future / Q-Bench

①[ICLR2024 Spotlight] (GPT-4V/Gemini-Pro/Qwen-VL-Plus+16 OS MLLMs) A benchmark for multi-modality LLMs (MLLMs) on low-level vision and visual quality assessment.
https://q-future.github.io/Q-Bench/
Other
251 stars 12 forks source link

no annotation json file for lldescribe data #12

Closed yeppp27 closed 10 months ago

yeppp27 commented 10 months ago

Hi! This work is very great. And I can't fine the json file for description annotation for lldescribe data. Thank you in advance~!

teowu commented 10 months ago

Hi~These LLDescribe GT Annotation are not planned to be released yet, and can only be tested via submission, as we plan to hold a competition based on this later. You may look at Q-Pathway dataset (a subset of Q-Instruct, our another project: https://huggingface.co/datasets/teowu/Q-Instruct, where qtype is null) and fetch some for validating your model.

Thank you for understanding and hope that helps.