Qwen_2_5_vl Object Detection Support?

[David-19940718]

Search before asking

• I have searched the Multimodal Maestro issues and found no similar feature requests.

Description

Does it support to object-detection fine-tune via Qwen2.5-VL?

Use case

No response

Additional

No response

Are you willing to submit a PR?

• Yes I'd like to help by submitting a PR!

[SkalskiP]

Hi @David-19940718 👋🏻 In general, you can already train Florence-2, PaliGemma 2 and Qwen2.5-VL for object detection, but your dataset would have to be converted to JSONL format to be loaded.

maestro-1.1.0 will significantly simplify this process. We will enable training directly from the COCO / YOLO dataset. And maestro will handle parsing internally. We will also provide traditional computer vision metrics such as mAP, which you will be able to use for training and benchmarking. You can read more about it here.

I am marking the issue as a duplicate, as we are already planning these features in the mentioned issue. Feel free to comment there if you still have questions.