YOLOv9 Training on Custom Dataset in Google Colab with Roboflow

Inside my school and program, I teach you my system to become an AI engineer or freelancer. Life-time access, personal help by me and I will show you exactly how I went from below average student to making $250/hr. Join the High Earner AI Career Program here 👉 www.nicolai-nielsen.com/aicareer (PRICES WILL INCREASE SOON)
You will also get access to all the technical courses inside the program, also the ones I plan to make in the future! Check out the technical courses below 👇
In this video 📝 we are going to take a look at the new YOLOv9 model. I'm going to show you how you can grab a dataset, export it into a Google Colab notebook and train a custom YOLOv9 model on your own dataset.
Colab Notebook: colab.research.google.com/dri...
If you enjoyed this video, be sure to press the 👍 button so that I know what content you guys like to see.
_____________________________________________________________
🛠️ Freelance Work: www.nicolai-nielsen.com/nncode
_____________________________________________________________
💻💰🛠️ High Earner AI Career Program: www.nicolai-nielsen.com/aicareer
⚙️ Real-world AI Technical Courses: (www.nicos-school.com)
📗 OpenCV GPU in Python: www.nicos-school.com/p/opencv...
📕 YOLOv7 Object Detection: www.nicos-school.com/p/yolov7...
📒 Transformer & Segmentation: www.nicos-school.com/p/transf...
📙 YOLOv8 Object Tracking: www.nicos-school.com/p/yolov8...
📘 Research Paper Implementation: www.nicos-school.com/p/resear...
📔 CustomGPT: www.nicos-school.com/p/custom...
_____________________________________________________________
📞 Connect with Me:
🌳 linktr.ee/nicolainielsen
🌍 My Website: www.nicolai-nielsen.com/
🤖 GitHub: github.com/niconielsen32
👉 LinkedIn: / nicolaiai
🐦 X/Twitter: / nielsencv_ai
🌆 Instagram: / nicolaihoeirup
_____________________________________________________________
🎮 My Gear (Affiliate links):
💻 Laptop: amzn.to/49LJkTW
🖥️ Desktop PC:
NVIDIA RTX 4090 24GB: amzn.to/3Uc7yAM
Intel I9-14900K: amzn.to/3W4Z5Cb
Motherboard: amzn.to/4aR6wBC
32GB RAM: amzn.to/3Jt2XVR
🖥️ Monitor: amzn.to/4aLP8hh
🖱️ Mouse: amzn.to/3W501GH
⌨️ Keyboard: amzn.to/3xUGz5b
🎙️ Microphone: amzn.to/3w1F1WK
📷 Camera: amzn.to/4b4Ryr9
_______________________________________________________________
Timestamps:
0:00 Intro
0:35 YOLOv9 Github
3:16 Code Setup
5:38 Inference on image Pretrained Model
9:40 Dataset Annotation & Export
12:37 Train Custom YOLOv9
14:33 Traning results
17:55 Infrerence on Image Custom Model
19:12 Outro
tags:
#YOLOv9 #ObjectDetection #customdataset

[@NicolaiAI]

Join My AI Career Program 👉 www.nicolai-nielsen.com/aicareer Enroll in My School and Technical Courses 👉 www.nicos-school.com

[@clayworld-zn8oz]

Hello!! It is an excellent video. I have a doubt. I have created a custom dataset and followed your instructions. It has worked excellent for an image but not for a video. Can you please let me know how it can be done to detect the objects from a video??? Please.. Thank you sir.

[@electrictechwithtaha]

Thank you soooooooooo much brother for your alot of effort in making this helpful video for us only you are the one which give us alot of info in this video and i understand your method again thank you soo much and keep it up.....love u

[@NicolaiAI]

It's my pleasure! Thank you so much for the kind words

[@balveersingh3051]

it's amazing to see how fast technology is evolving

[@NicolaiAI]

Yeah that’s for sure!

[@leosmi1]

yeahh, when we think yolo can't improve we get a new version.

[@ibnu7235]

thank youuu nicolai, amazing video

[@NicolaiAI]

Thanks you! Definitely check out the new v10 as well. Much better video

[@ninono12345]

Very good videos man, there are a ton of tutorials that do only the basics, but it's almost impossible find one that goes deep, so THANK YOU for posting your videos

[@NicolaiAI]

Thanks a ton for the nice words man! Really appreciate the support and exactly why I’m doing it

[@rozluthermeesala2202]

Hey!! I really love your video. May I know what did you do at 6:42 ? I got the same error you got and I am trying to resolve it. May I know where did you paste the image ? Thank you.

[@anningkyeireginald8453]

Sir i to use this on object defection where it will detect that this bottle is defected and give the areas where the defecction is say the label and vice versa too. Do you have video on that or how to implement that?

[@shahriarahmadfahim6457]

What is --close mosaic?

[@ronaldo-ir2iz]

at train custom model, how to specify the batch, and image? we have to refer from where?

[@heejins8907]

Thank you so much for this video. Is it feasible to train YOLOv9 on a custom dataset for different types of waste (paper, plastic, glass, etc.)? I will be the one capturing the images and annotating them.

[@NicolaiAI]

Yeah for sure! You can do that

[@MegaLenni2]

@NicolaiNielsen thanks for the tutorial. When using roboflow and the yolov9-c or yolov9-e model I get an AttributeError: 'list' object has no attribute 'view' in the train.py file line 168: File "/content/yolov9/utils/loss_tal.py", line 168, in pred_distri, pred_scores = torch.cat([xi.view(feats[0].shape[0], self.no, -1) for xi in feats], 2).split( can somebody help me solving this issue?

[@greatsarmad5147]

Awesome video as always man! I have a technical question since I'm new to this. How can I create a bounding box in a specific area and check if it contains anything? For instance, how do I draw a box over a parking spot on an image or video and ask the model to count cars within that box? I'm thinking of something similar to the line-crossing counting method, but applied to a whole box. My goal is to determine if something, not just what is recognized, is within a predefined area.

[@SquatsOfAwesome]

Supervision has a PolygonZone and PolygonZoneAnnotator class, which lets you mask out an area and filter for all detections within that area

[@NicolaiAI]

Thanks a lot for watching man! Appreciate it. Yeah you can use the zones from supervision. They have it all integrated in but you could also specify your own polygon zones in opencv and check if points are inside those

[@manurono7111]

Is there a way that one can get an annotation .txt file with the predicted image

[@SahilNegi2004]

i am getting this error File "/content/drive/MyDrive/yolov9/train.py", line 196, in train assert mlc < nc, f'Label class {mlc} exceeds nc={nc} in {data}. Possible class labels are 0-{nc - 1}' AssertionError: Label class 1 exceeds nc=1 in /content/drive/MyDrive/yolov9/dataset/data.yaml. Possible class labels are 0-0 any idea how to solve this error ?please

[@ronaldo-ir2iz]

"RuntimeError: DataLoader worker (pid 6246) is killed by signal: Killed."?. What does it means?

[@ronaldo-ir2iz]

can i know what to custom on data.yaml cuz in my data doesnt have valid data? train, 80%, test 20%

[@EgorPristavka-hn1to]

Does this model support OBB? If I'm not mistaken, 8.1 did, but what's about 9th version?

[@NicolaiAI]

Nope they are different. OBB is a feature only ultralytics have implemented on top of their yolo model

[@ReadyMedia-no]

Thank you! Are there currently any AI Edge cameras able to run this model? We are looking for a solution to count people in large auditoriums with 500 seats and 30M distance between Camera and furthest person.

[@NicolaiAI]

Thanks for watching! Don’t think any camera is out there for this model yet. I’m not too familiar with the available options out there for edge cameras running models out of the box

[@SAURSINGH]

You can look at a camera working with an Edge Computing Node like NVIDIA Jetson

[@muhammadrezaphahlevi1573]

I can't train my dataset using gelan-e, yolov9-c, and yolov9-e. Anyone can help?

[@minditon3264]

Does anyone know how to resolve this error File "/content/yolov9/utils/augmentations.py", line 248, in copy_paste l, box, s = labels[j], boxes[j], segments[j] IndexError: list index out of range

[@maazkhan4410]

i am also facing this error if you have solved it plz provide me the solution

[@user-do6jw9jq5g]

Николай, спасибо!

[@Zinou485]

Which version is better detection for you, 8 or 9?

[@YigalBZ]

As far a I recall, Roboflow doesn;t allow to download the weights file, right? What are my options if my applications require off line work on a local computer?

[@InfernalPasquale]

Build the model on your local computer

[@YigalBZ]

@@InfernalPasquale Can I do the training with Roboflow on my local machine?

[@InfernalPasquale]

@@YigalBZ if you mean train using your roboflow dataset then yes.

[@NicolaiAI]

You can follow my videos for custom training of yolo v8 and v9. You can directly export dataset and use that for training. Then u can export the model to your local machine

[@tenns-tu1yz]

nvidia-smi: command not found

[@XEQUTE]

cool video, I was thinking, it would amazing if you could do a video on yolov9's face recognition capabilities. I am working on a door access project and was thinking of using yolov9 for face recognition

[@NicolaiAI]

I have some videos on my channel with face recognition as well. You would normally use a bit different method and model for recognition compared to object detection

[@XEQUTE]

yes. that's what I was thinking@@NicolaiAI, but can object detection be combined with face detection in one model?

[@user-oh9ip6nn8t]

while training yolo9 on custom dataset, i am getting an error " AttributeError: 'list' object has no attribute 'view' " in line no 168 in loss_tal.py. can you please look into it and give me any solution?

[@NicolaiAI]

Looks like they have some bugs in the GitHub repo. Can u try clone the newest version? GELAN model should work

[@jacksonteo5787]

I am also experiencing the same error using yolov9-c

[@Fellixify]

Same error for me with v9. Gelan worked extremely well for me.

[@NicolaiAI]

@@Fellixify yeah they have to fix the bugs first

[@user-fw1yl8fd3d]

why no comparison to YOLO NAS?

[@NicolaiAI]

Will do

[@markrusso6678]

When I try to run the yolov9-c. i get this error AttributeError: 'list' object has no attribute 'view'

[@muhammadrezaphahlevi1573]

Same bro, I can't train my dataset using gelan-e, yolov9-c, and yolov9-e too

[@giuliobenocci5658]

Question: what type of stream does it need for a proper analisys?

[@NicolaiAI]

What do u mean by stream?

[@giuliobenocci5658]

@@NicolaiAI I mean rstp protocol with H265+ encoding or something

[@josephjesus19]

can you make a video about visual odometry for UAV? 🙏

[@NicolaiAI]

Don’t really have any footage for that but have a bunch of videos for visual odometry in general

[@leosmi1]

Is there a way to train in windows with intel gpu?

[@NicolaiAI]

Nope every framework is using PyTorch pretty much and only supports cuda. Also amd and mps but have never seen anyone use those. No reason not to go for nvidia gpu or mac if you prefer just doing cool stuff locally. I know you probably have an intel one already and that’s why. But they can be used to deploy models with openvino and get pretty good performance

[@leosmi1]

thanks @@NicolaiAI

[@akapholsaikong8062]

I followed the instructions but it shows an error saying IndexError: list index out of range.

[@NicolaiAI]

Where?

[@akapholsaikong8062]

@@NicolaiAI Traceback (most recent call last): File "/content/yolov9/train.py", line 634, in main(opt) File "/content/yolov9/train.py", line 528, in main train(opt.hyp, opt, device, callbacks) File "/content/yolov9/train.py", line 277, in train for i, (imgs, targets, paths, _) in pbar: # batch ------------------------------------------------------------- File "/usr/local/lib/python3.10/dist-packages/tqdm/std.py", line 1181, in __iter__ for obj in iterable: File "/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py", line 630, in __next__ data = self._next_data() File "/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py", line 1345, in _next_data return self._process_data(data) File "/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py", line 1371, in _process_data data.reraise() File "/usr/local/lib/python3.10/dist-packages/torch/_utils.py", line 694, in reraise raise exception IndexError: Caught IndexError in DataLoader worker process 0. Original Traceback (most recent call last): File "/usr/local/lib/python3.10/dist-packages/torch/utils/data/_utils/worker.py", line 308, in _worker_loop data = fetcher.fetch(index) File "/usr/local/lib/python3.10/dist-packages/torch/utils/data/_utils/fetch.py", line 51, in fetch data = [self.dataset[idx] for idx in possibly_batched_index] File "/usr/local/lib/python3.10/dist-packages/torch/utils/data/_utils/fetch.py", line 51, in data = [self.dataset[idx] for idx in possibly_batched_index] File "/content/yolov9/utils/dataloaders.py", line 656, in __getitem__ img, labels = self.load_mosaic(index) File "/content/yolov9/utils/dataloaders.py", line 791, in load_mosaic img4, labels4, segments4 = copy_paste(img4, labels4, segments4, p=self.hyp['copy_paste']) File "/content/yolov9/utils/augmentations.py", line 248, in copy_paste l, box, s = labels[j], boxes[j], segments[j] IndexError: list index out of range

[@VisionCLA]

@@akapholsaikong8062 Same error

[@razanalomari8242]

i have the same problem and it is during training , i edited the yaml file the same way as you did @@NicolaiAI

[@leosmi1]

I hate how dificult object detection is compared to image recognition.

[@NicolaiAI]

Depends on the task I’d say but segmentation is even more painful haha

[@viddeshk8020]

YOLO NAS VS YOLOV9

[@NicolaiAI]

For sure! But can probably already say yolov8

[@toyly2820]

You are speaking so fast as if you need to go somewhere urgently and someone is staring at you waiting till you are done. Give people some time to comprehend some steps man

[@user-mh8kb5ng9f]

I feel the same when I see this video series for the first time🥺, cuz English is not my first language. But when I go through all the YOLOv8 episodes, I can follow his words now. So, listen more maybe the only solution. 🤣

[@NicolaiAI]

It’s a trick to make you watch more videos. Jokes aside, you can slow down the video. I only get this comment 1 time/month so will not make sense to change the pace or flow of the videos

[@balarog7418]

​​@@NicolaiAI yes please don't slow down your pace, its perfect rn for most of us

[@NicolaiAI]

@balarog7418 thanks a lot man! Appreciate it

[@joycegoes266]

Jaante ho I met you because daddy did black magic on me. We met because of our problems. If daddy did not do black magic on me then I would be could not meet you. If I did not meet you then how you would be? Ask god know

[@abdelkarimlahmdani6793]

hi, i want to add tensorbord to this code how can i do it please can you help # Affichage des résultats à partir des journaux de TensorBoard %load_ext tensorboard %tensorboard --logdir {logdir}

[@freecode.ai-]

THIS IS SHOCKING!

[@NicolaiAI]

Shockingly good or? 😂

[@DarpaProperty]

Every AI/Vision video shocks the industry.

[@freecode.ai-]

Your video is great but it was just a meme statement. I have been meaning to ask, do you own ultralytics? It's like you are the face of the company when I visit the website.