Xuan-World / Mamba-YOLO-World

Mamba-YOLO-World: Marrying YOLO-World with Mamba for Open-Vocabulary Detection
GNU General Public License v3.0
40 stars 4 forks source link

Inference speed comparison #2

Open HaisongDing opened 1 month ago

HaisongDing commented 1 month ago

Mamba-YOLOs seem doding better than YOLO in terms of FLOPs and accuracy. How about their inference speed performances (e.g., Tensor-RT FPS)?

myb1314yxy commented 1 month ago

Thanks for the YOLO-World model integrated by your team, I would like to ask how I should use yolo-world to detect unknown categories in my own dataset, how to divide and make the dataset, and do I need to pre-define all known categories in yaml files?

Xuan-World commented 1 month ago

Thanks for the YOLO-World model integrated by your team, I would like to ask how I should use yolo-world to detect unknown categories in my own dataset, how to divide and make the dataset, and do I need to pre-define all known categories in yaml files?

For finetuning on your own dataset, you can refer to the docs of the original YOLO-World. The pipeline to organize datasets is the same. Also, you can try zero-shot inference as you wish.