BenjaminLi

@AyushExel has there been any progress on this?

I moved my data.yaml file into the root directory of the folder in an attempt to try solve this issue. I was able to get training working. For context, the...

I found that if you write everything as an absolute path from the root directory in colab, it seems to work fine. `!yolo detect val model=/content/drive/MyDrive/yolov8/runs/detect/train21/weights/best.pt data=/content/drive/MyDrive/yolov8/data.yaml` Try something along...

Hi, will this be merged soon?

Employ hyperparameter evolution: https://docs.ultralytics.com/yolov5/tutorials/hyperparameter_evolution Hope this helps!

@MyFirst905 I have not "solved it" but have a rough idea as to why this is the case. According to the paper: "With one output, the model will average multiple...

@heyoeyo thanks for the response. As you pointed out, it's not difficult to disable batch prompts altogether, but do you see a way to "selectively allow" prompting within batches? For...

I don't see why it wouldn't be able to accept prompts. Simply modify the code in `sam.py`. ``` sparse_embeddings, dense_embeddings = self.prompt_encoder( points=None, boxes=None, masks=None ) ``` You would probably...

@zzzyzh the original SAM uses multimask outputs to describe objects at different levels of detail. This repository seems to retrain SAM's multimask outputs such that it can predict semantically. No...

@zzzyzh minimal modifications need to be made on the code side. In mask decoder's `forward` function, you can find the following code: ``` masks, iou_pred = self.predict_masks( image_embeddings=image_embeddings, image_pe=image_pe, sparse_prompt_embeddings=sparse_prompt_embeddings,...