大脑爆炸了can i use it to auto annotation the image for create a dataset to train the custom Yolo v8 model?

[Mayuyu007123]

the labels be like this

[Mayuyu007123]

since i realize the YOLO-World + EfficientSAM and the hugging face demo resulted not the same, hugging face demo is better

[wondervictor]

Hi @Mayuyu007123, firstly, could you provide more details about your demo? Secondly, could you share your sample image? Thirdly, YOLO-World exactly can be used as an automatic annotator and an open-source project has involved it, see https://github.com/AILab-CVC/YOLO-World/issues/120. However, in my opinion, fine-tuning YOLO-World on your data with fewer epochs can obtain better performance on YOLOv8, you can refer to configs/finetune_coco

[ljyan93]

hey @wondervictor could you point us which model & config & script (with NMS) to reproduce the huggingface demo locally?

thanks in advance.