YOLO11: How to Train for Object Detection on a Custom Dataset | Step-by-step guide

[@qazijawad6451]

Great explanation bro.. Exact and to the point🙌

[@RichMiniön-r2m]

you guys are going to blow up cheers😉

[@Roboflow]

🤞🏻

[@ericlyons11]

One of the best videos on yolo. Thanks. I would love to see a similar video on training pose/key point data. Especially how to look at the training result metrics and optimize the training for the key points. Currently I'm having trouble training a custom key point dataset on normal objects that don't have skeletons. Yolo is finding the objects but the key points are not in the correct place.

[@Roboflow]

You mean YOLO11 pose?

[@ericlyons11]

@@Roboflow Yes. I'm trying to train a custom pose dataset to calculate the orientation between the keypoints. But currently I'm having trouble getting the keypoints to be detected in the correct place. Thanks.

[@구연산-y2l]

Informative video 🎉

[@abhishekfnu7455]

Thank you so much for this brilliant video.

[@Roboflow]

Thanks a lot for kind words!

[@ВиталийИщенко-ь4к]

I need detect near 10 object classes and work as fast as possible in real time, and second variant not fast but with maximum accuracy. Can you give me advice? may be specific yolo version?

[@tomaszbazelczuk4987]

great video as usual

[@sharathsrinivasan1]

Thanks for sharing,, very usefull... btw,, I love your linux prompt.. can you share the configuration for it?

nice video, i have a question that how to check a model on video in colab ? anyone knows or any tutrioal on it

[@hebertwattimena5170]

where u get the name of your workspace in 1:03:15

[@sofiacollado5807]

I'm working on using YOLOv11 for food ingredients detection, may I ask could Roboflow API host it or it is not supported for deployment, as I need the python sdk for inference calls?

[@Roboflow]

Yup! Roboflow can host the model and you can just run inference over the API.

[@arafatmuktadir]

36:04, if the batch size is 10 (not the epoch), the number of iteration would be 100 in order to go through 1000 images.

[@adriansupp8070]

Hi! Currently working on a computer vision project about vehicles. Can you share how do I deploy YOLO11 on Jetson Nano c100 (csi camera)? Is this possible?

[@Roboflow]

I recorded this tutorial some time ago: kzbin.info/www/bejne/jmrNpWunl8ScndE I’m not sure if it’s still up to date :/

[@repswithroscoe]

Can we please get a Flutter plugin for YOLO that is similar to flutter_vision but updated for YOLO 11?

[@Roboflow]

I'm not so good at mobile programming.... I only had a few classes with Android and Swift at university a few years ago. But it sounds interesting, can you give me some details, I can look around.

[@abhinavvyas3143]

how can i deploy my model to a security camera? and save all the results in a .txt file?

[@nguyenthidiemhuongk17ct53]

Suppose my dataset is in the form of yolo, if I have divided it into train, valid, test, and the image may have many labels, how can I balance it? Please help me

[@Roboflow]

Are you talking about class balance? That some classes are over/under represented?

[@nguyenthidiemhuongk17ct53]

@@Roboflow yasss, i want to balance the class:(

[@RaviShankerSingh-rt6ui]

Label issue: I’m currently working with millions of images across several classes and I'm looking for a way to automate labeling in YOLO format. Experiment: I tried training with a small dataset, I noticed that around 50 objects were missed to labeled due to low visibility or ignore by model. Advise: Do you have any suggestions for tools or libraries that could help auto-label unseen images from the internet? I’d appreciate any recommendations you might have!

[@Roboflow]

Really cool problem to solve. Can you tell me how many classes you need to detect? What is your task? Do you need to train the model or just annotate the data? If you need to train the model what is the expected speed? Does the model need to run in real time?

[@RaviShankerSingh-rt6ui]

@@Roboflow I'm developing a model focused on accurate measurements and advanced capabilities, typically handling 10+ classes per instance. We're looking to train it with YOLO to enhance object detection and key feature extraction. However, our dataset isn't annotated in YOLO format. Could you suggest any tools, libraries, or strategies to automate or streamline the annotation process, especially for large datasets?

[@usamasadiq1681]

Hi! my laptop has Cuda 12.4 installed will the pytorch with cuda124 work with Yolo11 or do I need to install the cuda 11.6 Toolkit? i am using WSL2.

[@Roboflow]

Version should match your hardware. So install PyTorch with 12.4 support.

[@vladkor5272]

I had the same problem as you. Solved it by installing cuda 11.6. This is the latest version of cuda supported by pytorch

[@PriyankBolia]

Can we use the models of yolo 11 with tensorflow js, I mean run in browser using JS or WASM, instead of python. Or alternatively is there something that can run the model without having to install each dependency on android or PC and expose some api endpoint or terminal commands.

[@Roboflow]

You can export YOLO11 model to TF.js docs.ultralytics.com/integrations/tfjs/#how-do-i-export-ultralytics-yolo11-models-to-tensorflowjs-format, but I’m pretty sure you can export is to ONNX and run in web runtime as well: onnxruntime.ai/docs/get-started/with-javascript/web.html.

[@uyendang3723]

Can I use COCO format label for YOLO11?

[@Roboflow]

Not directly. But you can covert your coco dataset to yolo format.

[@SwekeR4603]

day sorted🤩

[@ZeryanGuerra]

Awesome.

[@aneerimmco]

good tutorial :)

[@brianlamport7575]

I need help integrating my model with a website I built.

[@Roboflow]

What model are you using? What's the website you are building? Do you run the model on client side or on server?

[@brianlamport7575]

@Roboflow Thank you for responding. I built a small website of object detection that Identifies insect species. Prepared dataset using roboflow trained both internally and externally using yolov9. Currently it is just running on my laptop then later on I will try deploy it on an external server.

[@KammzTV]

Hi, does your ia do real-time detection? If so, how did you do it? I'm currently working on an object detection application and I need to use my phone's camera to detect objects in real time. Thanks

[@HalesAi]

Hello Peter, I have a small project I would like to discuss with you or any expert on your side to see if the project is feasible and also what it would take to get you to build. How can I reach you directly?

[@Roboflow]

Hi 👋🏻 you can try describe everything here. I’ll try to help in the comments. I don’t really do any consulting in DMs :/

[@Lanre4pf]

How can i use the trained model for real time detection

[@Roboflow]

We don’t have YOLO11 real time detection tutorial yet; but it is almost the same as YOLOv8. Take a look here: kzbin.info/www/bejne/h4ebZpiPhMdqnc0si=VvCvE7KPCrPb0k8A

[@raymond_lazer]

Rock N Roll !!

[@codures]

31:58 it's *relative path to absolute path*, not "global" 😉.

[@Roboflow]

Oh. I really can’t call it global?

[@codures]

​@@Roboflow you may call it any way you like, it's your call 😁. "global path" refers most times to the $path env var, in both windows and nix systems. Edit: very nice and complete video.

[@Geoffrey-8814]

From the video cover, I thought that YOLOv11 utilized a transformer architecture😂

[@Roboflow]

I was worried people may look at this cover and think so :/

[@rosemutegi8830]

Thanks for video, am stuck on how you tested the model using other data need you input; yolo detect predict source=/content/drive/MyDrive/assets/Test Model=/content/runs/detect/train3/weights/best.pt;!yolo detect predict source=/content/drive/MyDrive/assets/Test Model=/content/runs/detect/train3/weights/best.pt

[@Roboflow]

Did you got any error when running this command?

[@rosemutegi8830]

@@Roboflow Yes need help, training my data was okay but on using model to test getting errors