Update README.md

[0xSynapse]

Hi @jameslahm,

Ankan here. I am a Computer Vision Engineer at OpenCV. Thanks for YOLOv10; it's really cool with its new advancements.

Recently, I wrote a detailed article on YOLOv10 at LearnOpenCV.

I thought maybe we could add a readme button for the blog to help the community understand it better. As I did the same for YOLOv9.

Looking forward to it.

Thank you!

[jameslahm]

@0xSynapse Thanks a lot! We appreciate your efforts very much and we are very pleased to highlight your blog in the README. Additionally, we found some inconsistencies with our understanding and could you please help us on these? Thank you!

1. In this example and this example, we found that YOLOv10 fails to detect some small objects or objects in the distance, e.g., the person in the subway and the person in the distance in the left down corner. We previously provided some clarifications and suggestions here. Could you please set a smaller confidence threshold for YOLOv10? Thanks!
2. Is the AP₅₀ a typo in this table? The AP^val of YOLOv10-L and YOLOv8-M is 53.2 and 50.2, respectively, rather than AP₅₀. Thanks!
3. In this table, do you measure the inference speed and FPS for YOLOv10 using the pytorch model? Due to that in the non-exported format, e.g., pytorch, the speed of YOLOv10 is biased because the unnecessary cv2 and cv3 operations in the v10Detect are executed during inference, could you please set the export attribute of v10Detect to True before the measurement using the pytorch model like below? Thanks!

from ultralytics import YOLOv10

model = YOLOv10('yolov10n.pt')
model.model.model[-1].export = True
model.model.model[-1].format = 'onnx'
del model.model.model[-1].cv2
del model.model.model[-1].cv3
model.predict(...)

4. In the YOLOv10 – Benchmarks section, could you please also set the export attribute of v10Detect to True before the measurement using the pytorch model? Thanks!

Thank you very much! ❤️

[0xSynapse]

Hi @jameslahm, thanks for the suggestions,

2 - yes, it's a typo. It's fixed now.

1, 3, 4 - As this is an introduction blog, we are more focused on using the default parameters. We will write more blogs about this later when we experiment with the parmas. For now, I will add one subsection for each section, as you mentioned the issues and solutions.

I hope this will help!

Again, thanks for the suggestions.

[jameslahm]

@0xSynapse Thanks a lot! Could you please also replace the data of YOLOv9-C with its AP^val in this table? Thanks!

[0xSynapse]

@jameslahm , Sorry for the typo it's fixed now.

Rest 2 issues and solutions, I will add today itself.

Thanks for the help!

[0xSynapse]

@jameslahm updated your issues and suggestions.

Now I think we can add it to the README.

[jameslahm]

Thank you!