YOLOv10: Train a Custom Model and Run Inference on Live Webcam

[@NicolaiAI]

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

[@josetomasl.4611]

please create open source courses, it would help much more people, education should be free

[@NicolaiAI]

@@josetomasl.4611 have all the stuff available for free on my KZbin channel! Almost 500 videos. My courses are just structured, with help and other stuff like quizzes etc. creating 500 videos is not easy and takes 3-4 years. I have recorded videos getting back after New Year’s Eve at 3am in the morning. Of course I can sell some courses to make some money when I do it all for free. Bet all my money on you are and have been paying for education your whole life and everyone is. KZbin is the best platform for free education and I do my part

[@dakaradakaradakara699]

Hello Sir can we also embed this project with the custom dataset of single classification on roboflow?

[@IOSALive]

Nicolai Nielsen, You're the best! I subscribed because I love your content!

[@NicolaiAI]

Thanks a lot for the words man! Hope you will enjoy future videos too

[@IOSALive]

Nicolai Nielsen, This made me laugh so much! Thanks for sharing!

[@NicolaiAI]

Glad you enjoyed it!

[@HemangJoshi]

Best tutorial ever

[@NicolaiAI]

Hey man, thanks a lot for watching!

[@felixkuria1250]

Thank you so much for sharing this. I want to use the same in fish fingerlings tracking and counting, any advices on how I should go about it

[@FlyWithSergio]

Amazing tutorial. Thanks a lot.

[@NicolaiAI]

Thanks a ton man! Definitely check out the new video with training in Ultralytics as well. It’s added now into the framework

[@alberthades202]

the consumption seems better, but actually it doesnt give me the thought like disruptive

[@NicolaiAI]

100% correct!

[@egamomusu]

how to save yolov10 training process in Google Drive, in order to be able to resume the training if it was interrupted and i want to continue the training from the last checkpoint?

[@danielcu96]

Awesome video Nicolai, yolov10 is great news for the space. I have been doing some tests with the models exported to OpenVino and i can get some pretty decent inference time results on CPU with this new family of models, a good 20-30% improvement w.r.t the equivalent yolov8 models.

[@NicolaiAI]

Thanks a lot for watching! And also sharing your insights and test. Thanks man

[@danielcu96]

@@NicolaiAI Thanks to you for the informative videos, your channel is awesome. Here are the simplified results from the tests using the c++ OpenVino inference engine on specific target hardware (Intel Core i5-1135G7 - 8 GB RAM). Yolov10n: 25.8471 ms (38.68 fps) Yolov10s: 70.372 ms (14.21 fps) Yolov10m: 185.095 ms (5.40 fps) Yolov10l: 376.334 ms (2.66 fps) I strongly believe this preliminary tests show that yolov10 models could be deployed in real-world, real-time applications where the only available hardware for inference are Intel CPUs, specially for the small (nano to medium models) and/or if you further explore the quantization techniques available within the Ultralytics repository or the NNCF tools from OpenVino.

[@raymondmichael4987]

Will come back to this 😊

[@NicolaiAI]

Definitely do!

[@nicolasgauttier9199]

Thanks for this useful tutorial. In the last part of the video for the inference, you don't put the model in eval mode 'model.eval()' , and you don't use 'with torch.no_grad()' for the inference. Is it useless in your case?

[@akshaygn4494]

Couldn’t able to import Yolov10 from ultralytics in pycharm

[@NicolaiAI]

Pip install the yolov10 repo as I do in the video. It’s not supported by Ultralytics own package yet

[@akshaygn4494]

@@NicolaiAI thanks. It worked!!

[@dakaradakaradakara699]

I want to request on how to do this project on classification sir, please.

[@jxuuls7880]

Why do a whole bunch of boxes show up on my object detection?

[@yasin6904]

Would it be possible to label events? For example, a highway camera and labelling where it's should label it as "traffic_congestion"?

[@NicolaiAI]

Yeah that would just be classification if u want to assign a label to the whole image

[@ragnarok9131]

Nice video!

[@NicolaiAI]

Thanks mate!

[@aryehrotberg]

Thanks! Does YOLOv10 have segmentation model as well? I haven't seen on the official website.

[@NicolaiAI]

Nope not yet

[@peteolesen265]

I live on a busy street and would like to do traffic counting. I would like to track pedestrians, bicyclists, cars, buses, etc. I am a newbie but would anyone recommend a specific wireless web camera? I am thinking of installing on a tree and looking for a solar recharge option. I would also like to track things 24/7 so night capability is needed. Just wondering what camera(s) and features I should look for. Just started to do some stuff in Python, but could be an interesting project. Thanks very much.

[@vasanth4599]

hello, i have get error (cannot import name 'YOLOv10' from 'ultralytics') and i install in local my ultralytics==8.2.48 version

[@FlyWithSergio]

Think you might have to change v8

[@vasanth4599]

@@FlyWithSergio why?

[@santhoshk2883]

I got error attribute error that v10 detectloss please help me to solve

[@tuzik8692]

Why in logs at 15:15 wrote that using yolov8. And a little bit later will be download yolov8n??

[@renpingol0]

TypeError: unhashable type: 'numpy.ndarray' how do i fix this error? i just need it with my project its due on this..

[@DarhanZhaksalykov]

hello, i have error (cannot import name 'YOLOv10' from 'ultralytics')

[@NicolaiAI]

Did u pip install the yolov10 GitHub repo as I do in the video?

[@DarhanZhaksalykov]

@@NicolaiAI yes

[@NicolaiAI]

@DarhanZhaksalykov locally or in colab? In colab it should work. Locally it might be because u already have another version installed of Ultralytics

[@jaimeabug9090]

can you do it from tge very first step? im new, i need this for my school proj

[@NicolaiAI]

This is from the very first step of generating a dataset and training a model and inference

[@jaimeabug9090]

@@NicolaiAI the cv2 said could not be resolved

[@fishfooties9850]

@@jaimeabug9090 if cv2 module is not recognized try reinstalling cv2, or install it in conda if ur using python

[@trendcip]

Hello bro Selam Aleyküm! i want custom dataset with yolov10 but some mistake! from ultralytics import YOLOv10 ImportError: cannot import name 'YOLOv10' from 'ultralytics' (C:..\Python\Python312\site-packages\ultralytics\__init__.py)

[@NicolaiAI]

Did u install yolov10 GitHub repo? Ultralytics doesn’t work yet

[@trendcip]

@@NicolaiAI No, I don't know how to do this.

[@NicolaiAI]

I show that step in the start of the colab notebook

[@trendcip]

@@NicolaiAI I have this problem when I start Realtime. I use Visual Studio.

[@NicolaiAI]

@trendcip make sure you have uninstalled Ultralytics and then install yolov10 as I do in the colab notebook

[@akshaygn4494]

How to display confidence score

[@NicolaiAI]

U can extract that directly from the results variable as well

[@akshaygn4494]

@@NicolaiAI how can I display it in the box?

[@DzBrain]

How can i contact you sir

[@NicolaiAI]

Have all my social media’s in the description

[@profapps]

how to installl v10

[@NicolaiAI]

You can use Ultralytics now! But I show how to install v10 here in the video but also have a new one for Ultralytics

[@Dani-h5g]

I think I love you

[@NicolaiAI]

Love u too

[@SuperYtc1]

This shit is so complicated

[@TalikkaMuhib]

I am facing these errors in training cell! TypeError: unhashable type: 'numpy.ndarray' TypeError: Caught TypeError in DataLoader worker process 0.

[@eliottvalette2453]

i'm facing it too, have you solved it ?

[@taliqamuhib8968]

Yes I resolve the problem

[@TalikkaMuhib]

@@taliqamuhib8968 !pip install albumentations==1.4 RUN THIS BEFORE TRAINING

[@renpingol0]

@@taliqamuhib8968 did you guys already fix it? and how?

[@juanjosevilca9513]

@@taliqamuhib8968 how?

[@renpingol0]

how do i fix this error nicolai? TypeError: unhashable type: 'numpy.ndarray'

[@juanjosevilca9513]

same :,c