OpenAdaptAI
Implement UGround
Feature request
We would like to incorporate UGround for identifying regions of interest:
https://x.com/ysu_nlp/status/1844186560901808328
Inference code here: https://github.com/boyugou/llava_uground/
Motivation
As an alternative to segmentation / description.
After reviewing the integration requirements for UGround, I estimate that the implementation will require approximately 2 to 4 days of work, considering the complexity of adapting the inference code from https://github.com/boyugou/llava_uground/ and ensuring proper testing and documentation. Therefore, I suggest a bounty range of $300 - $600 for this task. Looking forward to your thoughts!
Thank you @aybanda ! Can you please clarify your proposal? Since UGround requires CUDA and OpenAdapt is designed to run locally, I believe this will require implementing a separate UGround server and deploying it. What do you think about fleshing out what other steps are involved so that we can narrow it down? Please also feel free to message me on Discord with your reply.
got it @abrichr sent you a request in discord, I will send you detailed proposal there.