Михаил

I have prepared a draft that can go to the final: To run without faster-coco-eval (if it is not installed), run pycocotools \ lvis ```bash YOLO_AUTOINSTALL=false yolo detect val model=yolo11n.pt...

@Laughing-q What are my next steps?

> @MiXaiLL76 can we remove the older pycocotools from pyproject.toml? I'd like to keep dependencies to the minimum required. Of course, for this you need to replace https://github.com/ultralytics/ultralytics/blob/aabd0136ec40223cf423847635a2a4de95bba63d/ultralytics/models/yolo/segment/val.py#L265 with other...

> Thank you for the suggestion. We'll consider integrating your library for mask encoding, ensuring compatibility with numpy versions. If you can prepare a draft for conditional use, that would...

> Thank you for the suggestion. Please proceed with preparing the draft to use your library's function, ensuring compatibility with numpy versions. Once ready, we can consider removing pycocotools from...

> Thank you for the suggestion. We can consider replacing pycocotools with your library once it supports numpy>2. Meanwhile, feel free to prepare a draft for conditional use of your...

> Thank you for the update. We'll review the latest version and consider integrating it into our workflow. What are the plans for further implementation? How can I help?

> @MiXaiLL76 will you have time to check this out? 🦩 Hi! Thanks for the issue! I'll look into it as soon as possible and get back with a fix!...

https://github.com/MiXaiLL76/faster_coco_eval/pull/40 In general, I started a virtual machine with windows and checked everything in anaconda. Made the necessary corrections and now everything works. For macos, I made a "blind fix",...

https://github.com/MiXaiLL76/faster_coco_eval/pull/41 Regarding another bug, I wrote here https://github.com/Lightning-AI/torchmetrics/pull/2750#issuecomment-2403450693