Learn What Is Introduced in YOLOv10 | YOLOv10 Paper Explained

YOLOv10: Real-Time End-to-End Object Detection
Paper: arxiv.org/pdf/...
YOLOv10, developed by researchers at Tsinghua University introduces a novel approach to real-time object detection. This version addresses deficiencies in both post-processing and model architecture found in earlier YOLO versions. By removing non-maximum suppression (NMS) and optimizing various model components, YOLOv10 achieves state-of-the-art performance with significantly reduced computational overhead. Extensive experiments show its superior accuracy-latency trade-offs across multiple model scales.
#computervision #objectdetection #yolov9 #yolov8 #yolov10

[@rajmeetsingh1625]

Great information. Can you please make one video on how to get the inference time graph comparison with yolov8, yolov9, and yolov10 using the new features of yolov10?

[@CodeWithAarohi]

Sure, Soon!

[@sozno4222]

I love your channel. Really great stuff. If you can, I suggest buying an external microphone. Improved sound quality would do wonders to improve the quality of the videos.

[@CodeWithAarohi]

Thanks, will do!

[@mdk1171]

thank you can't wait to see you working on it

[@CodeWithAarohi]

Hope you like it!

[@cyberhard]

Nice. Thanks for the video. I didn't realize v10 was here. Or soon will be once it is incorporated into Ultralytics.

[@CodeWithAarohi]

Glad you found it helpful!

[@SaiSira]

🎯 Key Takeaways for quick navigation: YOLOv10 introduces a feature called "NMS free training" to avoid duplicate bounding boxes for the same object, reducing postprocessing time and computational resources. Spatial Channel Decoupled Down sampling in YOLOv10 separates spatial and channel operations to make downsampling more efficient, using pointwise and depthwise convolutions. Rank Guided Block Design in YOLOv10 adjusts model stages based on redundancy levels, improving efficiency by allocating compact inverted blocks where necessary. Lightweight classification heads in YOLOv10 are designed to be efficient in assigning labels without compromising accuracy. Made with HARPA AI

[@daniell9062]

I feel like the performance of recent YOLO models are similar anyway. I guess not having NMS may be nice for mobile devices or when computational resources are limited, at slight cost of performance?

[@rishabhjangid5575]

Can we convert yolov10 custom trained model to quantised tflite model

[@vinitsarode908]

you didn't explain how the method is NMS free. Is there any other video for this explanation?

[@CodeWithAarohi]

NO, I don't have any other video on this.

[@RightlyFree]

indeed it's very helpful, thank you very much Aarohi

[@CodeWithAarohi]

My pleasure 😊

[@KiranBatool-e6b]

Is anyone using this model? I'm using yolov9 model but its not providing any beneficial results same as yolo8l.

[@allanjobs3595]

can you make a details video how can I upgrade any yolo model to upgrade version for a project ? and also how can I we customize yolo model ? I'm studying ai now but facing problem for that , Lack of this type of tutorials videos in youtube , that will be help me a lot

[@CodeWithAarohi]

Sure, Soon!

[@MahmoudHashem-fh7ie]

Thanks for your effort

[@CodeWithAarohi]

It's my pleasure

[@rajmeetsingh1625]

Dear Maam, please suggest to me the Deep learning model and a way to detect the cursive Hindi character from the image and rewrite it in normal Hindi characters as output. I want to use Yolo. Could you suggest the methodology or any link? How do I label the Hindi character as Hindi text output?

[@CodeWithAarohi]

You can train yolo model on dataset which have hindi characters. For this- collect images of all hindi characters, annotate them and then train your model on it.

[@rajmeetsingh1625]

@@CodeWithAarohi thanks but I am confused how to annotate them . How do I add labels in Hindi ( अ, ए, ) any annotations software?

[@khalidamiralam6469]

Hello, great work. Can you plz make a video on text to image to video, and explain the python code as well

[@CodeWithAarohi]

Yes, sure

[@icon_chatt]

maam wanted to know from where did you get the links of the pt file for yolov10 for training

[@CodeWithAarohi]

You can download pretrained weights from here:github.com/THU-MIG/yolov10/releases

[@icon_chatt]

​@@CodeWithAarohi thank you maam

[@jhicinternational]

fantastic !

[@CodeWithAarohi]

Glad you like it!

[@RAZZKIRAN]

thank u madam, great content

[@CodeWithAarohi]

You are most welcome

[@pihao-r2w]

Can I ask you a few questions through other instant messaging apps?

[@CodeWithAarohi]

You can email me at [e-mail]

[@sush5651]

maam try running yolov10 on jetson nano.will it work?

[@CodeWithAarohi]

YOLOv10 is based on ultralytics and I was not able to run ultralytics on jetson nano. But using DeepStream, I was successful in running ultralytics models (yolov5, yolov8) on jetson nano. You can try the same for yolov10. Run yolov10 using DeepStream on Jetson nano.

[@sush5651]

@@CodeWithAarohi maam it would be very helpful if u implement this pls maam

[@aiforeveryone]

Great

[@CodeWithAarohi]

Thanks!

[@tilkesh]

🙏

[@drivemy5128]

thank you

[@CodeWithAarohi]

You're welcome

[@patis.IA-AI]

Thanks

[@CodeWithAarohi]

Welcome

[@daniell9062]

Damn you are faaaaaaaast

[@CodeWithAarohi]

Thanks! Just trying to keep up with the pace :)

[@DeepakYadav-jc8mo]

While I appreciate your attempt, your walkthrough is extremely superflous information, borderline redundant and offers very less in-depth info. But pls keep on going! Take this as a gradient update step into your learning!

[@CodeWithAarohi]

Thank you for your feedback and encouragement!

[@tikendraw]

I thought of yelling "OKAY MADAM🫡" in response to your "OKAY?" 😂

[@CodeWithAarohi]

Haha, that's funny! 😆 Glad my "OKAY?" got such a spirited reaction! Thanks for the laugh.