Train Yolov8 object detection on a custom dataset | Step by step guide | Computer vision tutorial

In this video I show you a super comprehensive step by step tutorial on how to use yolov8 to train an object detector on your own custom dataset!
Code: github.com/computervisioneng/...
Step by step tutorial on how to download data from the Open Images Dataset v7: bit.ly/46Ue4kC
🎬 Timestamps ⏱️
0:00 Intro
0:42 Data collection
5:04 Data annotation
19:49 Dataset structure
30:08 Train yolov8
50:15 Test yolov8
1:04:09 Outro
🌍 Community 👥
Join our Discord server: / discord
Support me on Patreon: / computervisionengineer
Hashtags #
#yolov8 #yolo #objectdetection #computervision #computervisionengineer #computervisiontutorial

[@ComputerVisionEngineer]

Did you enjoy this video? Try my premium courses! 😃🙌😊 ● End-To-End Computer Vision: Build and Deploy a Video Summarization API bit.ly/3tyQX0M ● Hands-On Computer Vision in the Cloud: Building an AWS-based Real Time Number Plate Recognition System bit.ly/3RXrE1Y ● Machine Learning Entrepreneur: How to start your entrepreneurial journey as a freelancer and content creator bit.ly/4bFLeaC All my premium courses are available to the Computer Vision Experts in my Patreon. 😉 www.patreon.com/ComputerVisionEngineer

[@johnmcook1]

did u aimbot the alpaca ?

[@ComputerVisionEngineer]

@@johnmcook1 😂 I did not

[@semtex6412]

@@ComputerVisionEngineer hi Felipe, will this wonderful tutorial work with the current YOLOv8 version and its dependencies? or we need the exact versions you used in this video? will there be errors? compatibility issues?

[@sreerajpainat]

,

[@user-zh1zo8lw2u]

the fact that you have shown how to do it in both colab and locally is just amazing. you are sooo underrated. please don't stop making content. your content is sooo awesome

[@gh0stofverdansk]

Been working something similar for a hackathon in my freshman year, first thought was to build a CNN from scratch (with no prior knowledge in computer vision) , then came across this video, you are a lifesaver man.

[@eleinaedelweiss6215]

thank you my man you wouldnt believe how much gratitude i have for this video and the fact you even showed how to run it locally as well. thanks man i reallly mean it.

[@odostanley1688]

I sincerely appreciate your indepth turial. It kicked started my journery in Computer Vision.

[@digitalchowder]

Thank you so much for your help in this video. The way you broke down each step of the process helped way beyond everything else I have found. Too many demonstration videos assume too much. You didn't. Thanks, again.

[@ComputerVisionEngineer]

You are welcome! Glad you find it helpful! 😃🙌

[@khalilsabri7978]

great work and thanks for the tutorial , you're the first youtuber I have seen answering everyone's comments . Continue the great work !

[@ComputerVisionEngineer]

Thank you! 😊🙌

[@harshans7712]

Your videos are really simple and shares good amount of knowledge, as a beginner your videos always helped me a lot

[@narenderrawal]

this is the most easy to follow tutorial on customized objects and I could train it myself for traffic lights. Thank you so much..bless you!

[@ComputerVisionEngineer]

Thank you! Glad it was helpful! 😃🙌

[@obensustam3574]

I really like watching your tutorials, you have great energy 🥳 and a good way of teaching the concepts 🤓

[@ComputerVisionEngineer]

😃💪💪💪

[@user-cw3vo8hv5u]

The video is of high quality, simple and practical, and avoids a lot of boring and difficult computer knowledge. As a beginner, I feel that I have received new knowledge, thank you.

[@ComputerVisionEngineer]

Thank you!! So glad you enjoyed it and that you found the content helpful! 😃🙌

[@coreytaylor5406]

Dude you are great ! At the begining i was struggling with your accent but it ended up well. I learned new things from you, such as data annotation, implementing YOLOv8. I am grateful to you.

[@ComputerVisionEngineer]

😃 Hey, thank you! Glad the videos are useful! I am now uploading subtitles in all my recent videos, to help people understand my accent. 😅💪

[@adipurnama_workshop]

this is the video I'm looking for. Thank you. Your accent make English clearer for me, so the lesson is very clear.

[@ComputerVisionEngineer]

Cool! Glad to hear you find the video useful! 😃🙌

[@MidoRaff]

I rarely comment on any videos, but yours was really one of a kind, love your way of explanation and teaching, wishing you well and truly thank you for your efforts!

[@ComputerVisionEngineer]

Thank you for your support! 😃🙌

[@srgberni]

Thanks a lot for this tutorial! I have spent hours in other tutorials that ended up being a copy paste of steps in Jupiter, but you went the extra mile to show and help us understand the whole process in our own computers. This is really good work and is helping us learn a lot. Thanks heaps!

[@anushkagupta1489]

Hey I am working on the same project and I have some problem while training the model. So can you help me out. Can you please send me your number so that I can ask my doubt as I have my hackathon tomorrow and its really urgent

[@rc3015]

the best video i have found on how to train yolo v8 thanks a lot

[@DearAi1]

youre so fluid coherent and elaborate 1 hour felt like 1 minute

[@ComputerVisionEngineer]

😊 Thank you so much for your kind words! Glad you enjoyed the video! 🙌

[@alperenkeser]

Superb tutorial! The best I've seen yet.

[@syirrus9845]

Thank you for taking the time to make this video.

[@anon-he4pg]

Thank you for this channel, your content is very educational and moreover you seem to be a nice person!

[@ComputerVisionEngineer]

😃 You are welcome! It is such a pleasure for me, I have so much fun working on this channel! Thank you for your kind words! 😊🙌

[@parsundback1648]

Hi! This was a great intro and really good explanation and it helped me get going with my 'recognize bees' project 🐝 Thanks for sharing your knowledge and you now got a new subscriber ;-)

[@ComputerVisionEngineer]

😃 Awesome! Glad the video was helpful. Best of luck in your bees project! 🐝🙌

[@sanjoetv5748]

I really love this channel he explains everything that you needed to know, 1 hour long video is not boring for a good teacher but 15 min. long video is too boring if the teacher is not good. keep on posting I am learning a lot from this channel

[@ComputerVisionEngineer]

Thank you for your support! 😃🙌 I will keep on posting! 💪

[@laurentchivot6528]

Hello and many thanks for this great tutorial , i have one question maybe not covered...when you annotate your pictures you don't tell about picture size harmonisation. Is it compulsory for better results ?

[@yuribelentani]

the best explanation about how to apply Yolo V8. thank you! Your video helps me a lot

[@ComputerVisionEngineer]

Thank you! Glad the videos are useful! 😃🙌

[@AnasHawasli]

This is so underrated! Great video

[@code___f1]

This is fantastic! Excited to give this a shot-appreciate it, mate!

[@ComputerVisionEngineer]

😃 You are welcome!

[@code___f1]

@@ComputerVisionEngineer do you have any good resources or suggestions for taking this trained model and then doing object counting in an ROI?

[@TheUnshifted]

Amazing Tutorial, After days of trying to figure vision models out with AI help and videos, your video was the only thing that managed to put me on the right track, I only ran into one issue but managed to figure it out pretty easily, Now I have this thing producing semi supervised datasets for further training, I Subscribed You Definitely deserve more attention. I didn't have to re watch the video, overly pause it to understand what is going on, Super easy to follow along, Overall one of - if not the best guide/tutorial videos out there! Thanks For the knowledge my guy, God bless!

[@ComputerVisionEngineer]

Thank you for your support! Glad the content is helpful! 😃🙌

[@tommasyu5136]

As a cv grad student，i wanna say thanks for your tutorial！！！helping me a lot

[@ComputerVisionEngineer]

😃 You are welcome! I am glad it is helping you, Thomas! 🙌

[@tommasyu5136]

@@ComputerVisionEngineerI appreciate it! Having a nice day ! My bro!

[@chesswithbill]

Spectacular tutorial, sir. Thank you.

[@digvijaysingh1285]

Thank you very much for the amazing tutorial please keep making these videos no such material is on KZbin very very helpful for each aspiring student

[@ComputerVisionEngineer]

😃 Thank you! I am glad you find it useful. I will continue posting similar videos! 😄🙌

[@digvijaysingh1285]

Please make custom dataset video object detection yolov8 code complete explanation - ( the last code in this clip)

[@ComputerVisionEngineer]

Do you mean the inference? Sure, I can do that. 🙌

[@digvijaysingh1285]

@@ComputerVisionEngineer yes please thank you sir

[@metehanguzel9081]

thanks, i've graduated thanks to you :D

[@steel-r_ua]

I need a thing to do and for that I need to learn ML, and you have so many interesting videos! They all are so tempting! 👍

[@anushka.narsima]

Thank you so much for covering different ways to do the job!!!! I've been struggling over the past 2 weeks bcuz every tutorial showed a single method and I kept running into some kind of error each time and had to redo the whole process so many times and downloaded a bunch of useless stuff >.

[@ComputerVisionEngineer]

Yeah I hear you, I've been there! I am happy you found the video helpful! 🤠🙌

[@RIDWANLEKAN-og2yx]

You just gave me a comprehensive insight about annotation of image. Thanks

[@ComputerVisionEngineer]

You are welcome! 😃💪

[@fatmanursefer1482]

You are my hero man!!

[@RohanAjmera-vw6mo]

very good tutorial! Loved it

[@shijiqiao3448]

Abosolutely helpful! So understandable and thank you so much!!!

[@ComputerVisionEngineer]

You are welcome! Glad it is useful! 😃🙌

[@bilalshahid7494]

Amazing tutorial! Can you please tell how can I determine exact values accuracy, precision and f1 score on my testing data?

[@carlitosvh91]

Great video. Thank you very much

[@davidpoling7748]

Thanks! This is by far the best guide I've found for this process

[@ComputerVisionEngineer]

Thank you! Glad you find it useful! 😃🙌

[@zeerakkhan274]

Great indepth explanation of the entire process. 👍👍

[@ComputerVisionEngineer]

Glad it was helpful!! 😃💪

[@dalecioustalk9964]

I am an aspiring AI engineer that is currently taking Electrical engineering in college. I am thankful to have found your channel today. I can now binge watch quality computer vision tutorials. 🎉🎉🎉

[@ComputerVisionEngineer]

Hey, that is so cool, my background is on electrical engineering too! ⚡I am glad you find my content useful. Happy binge watching! 😃🙌

[@ubuntulinux81]

@@ComputerVisionEngineer me too, i have question how can i use yolov8 in linux terminal

[@ComputerVisionEngineer]

Hey, I explain how to train a model using yolov8 from the terminal in this video, go to minute 41:38! 🙌

[@ubuntulinux81]

@@ComputerVisionEngineer it gives me so many errors while using in ubuntu terminal how can prepare from dataset in ubuntu

[@ComputerVisionEngineer]

have you installed the package in a local environment? what Python version are you using? what errors do you get?

[@Taki7o7]

Great Video for beginners ♥ A video about training an actual useful dataset like escooters and escooters + driver would be cool. If not possible with training, do it with code, like checking if a person and escooter box intersect, etc. And detecting multiple people standing on one scooter and stuff like that. I mean i love alpacas, but i cannot think about a usecase for that or some video to build on top of the previous ones. Just a suggestion to make it more interesting as real projects could be a great YT series :) Greetings from Vienna

[@ComputerVisionEngineer]

Thank you for your recommendations! I will try to make a more practical use case of object detection. Cheers! 😃🙌

[@Taki7o7]

@@ComputerVisionEngineer thanks for your answer, i hope you did not misunderstood it, you have great in-depth videos of course and i love them. I just meant to give beginners a more guided journey in form of a series with more practical use case/s. It might still sounds weird and like critique, but it's really just an idea with my chaotic english skills :D Greetings

[@hsnhejazin1177]

Thanks for this comprehensive tutorial. What if the goal is to add an additional class to the available classes in YOLOV8. Particularly, annotation tool starts annotating from label number 0. However, class 0 is available in corresponding classes in YOLOV8. does CVAT allow to label from the last index of COCO dataset to add a new class to the available ones in YOLOV8?

[@WelcomeToMyLife888]

Thanks for the video! Keep at it brother!

[@ComputerVisionEngineer]

😄🙌

[@andreapertosa5022]

This is THE HOLY GUIDE. Thanks!

[@ComputerVisionEngineer]

Glad you enjoyed it! 😄🙌

[@MsRAJDIP]

Thank you, you explained so well and detailed.

[@ComputerVisionEngineer]

You are welcome! Glad you enjoyed it! 😃💪

[@lefrancis5553]

THANK YOU FOR THIS EXCELLENT TUTORIAL

[@ComputerVisionEngineer]

😃🙌

[@user-zs6pi7gm6i]

Really helpful Video for understanding the procedure and I admire the simplicity! Can you please explain how to test with an image and display the output with the bounding boxes and class label?

[@ComputerVisionEngineer]

Thank you! Take a look at the script predict_video.py, I draw the bounding box and the text on top for all frames of the video. 🙌

[@haroldjoerges5766]

You have the magic stick, really god teaching

[@ComputerVisionEngineer]

😄 Thank you! Glad you enjoyed it! 🙌

[@kishorem5530]

Thank you sir, you are the best ...Keep up the good work love from india

[@ComputerVisionEngineer]

Thank you so much for your support! 🥰🙌

[@AdautoTDFNeto]

Amazing tutorial and explication, +1 subscriber

[@kientranasia]

Thank you for your sharing

[@derekboase9996]

Thank you for the great video! Would love for you to dive into the fine-tuning and optimisation for this model. Do you have anything like that? For example, improving results using statistics of the ground truth images, hyper parameter tuning, etc...

[@ComputerVisionEngineer]

Fine tunning and optimization would be a very good topic for a future video. 🙌

[@cattnation6257]

@@ComputerVisionEngineer did you make one

[@MBKerschdorfer]

You are awesome, thank you

[@Hm-bz4ps]

Best video I have seen! Keep it up!

[@ComputerVisionEngineer]

Thank you!

[@chandrakalagowda3129]

Thank you! for the detailed video.

[@ComputerVisionEngineer]

You are welcome! 😃🙌

[@vivekpatel009]

Hi, nice tutorial. I really enjoyed it. Is it possible for you to make a multi-label classification tutorial using DL algorithms?

[@ComputerVisionEngineer]

Hey, thank you, I am happy you enjoyed it! 😊 Sure, that is a great suggestion, I will try to do a video about that soon! 😃💪

[@abhishiktpeters2948]

Thank you so much.

[@midhunbose8228]

Gud job brother❤

[@CaioSantos-rh7yk]

nice man, i found a incredible content thank u so much!!

[@ComputerVisionEngineer]

You are welcome! Thank you for your support! 😃🙌

[@pyaephyo3633]

Thank you very much for sharing and explanation, Sir.

[@ComputerVisionEngineer]

You are welcome! 😊🙌

[@Nj-hw2cv]

Oh man you nailed the tutorial 💯

[@ComputerVisionEngineer]

😎💪 Glad you enjoyed it! 🙌

[@sayvan7585]

Muy buen tutorial amigo

[@rajashehryar2002]

Thank you soo much Sir

[@shwetabhat9981]

You are just AMAZINGGG !!! sir . Thank you so much :)

[@ComputerVisionEngineer]

You are welcome!! Thank you for your kind words. 😊🙌

[@subinlebow963]

thank you for the amazing tutorial! I was able to train YoloV8 on a custom data, but I was wondering how I can continue to train it using weights from previous training sessions. In other words, how do i make the second training session include the knowledge it gained from the first training session?

[@rma_sans4314]

Hi! I just came across you're tutorial and I love how detailed it is! But if I may ask, do you know how one would do transfer learning in YOLOv8?

[@ComputerVisionEngineer]

Hey, in order to do transfer learning from a model trained in the coco dataset you should create the model like this: model = YOLO('yolov8n.pt') 🙌

[@Aayushi1012]

@Computer vision engineer Hi, I was referring your video for one of my projects, it really crisp, clear and quite understandable. I really liked the way you teach! I had a doubt that can we export YOLOv8 to .h5 format?

[@ComputerVisionEngineer]

Hey, glad you found it useful. 😊 There may be a way to convert the yolov8 model into a keras (h5) model. Try to search for a model converter from Pytorch to keras. 😃🙌

[@sizeli7120]

This tutorial helps a lot, thanks~

[@ComputerVisionEngineer]

You are welcome! I am glad you find it helpful! 😃🙌

[@babcidy7963]

Hey, thank you very much. this was helpful. But when it’s done training, in the detects folder, the images have no bounding boxes. Please advice

[@micocoloso1621]

Your videos are very helpful! Thank you so much! However I have a question, If i want to do a multiclass object detection, do i need separate training folders or do i just put every image with different classes in one folder?

[@amarnathreddysurapureddy9111]

Hi Felipe, Ultralytics using any standardscale for detecting object. Original images to standard format

[@TheTechCricketer]

Super helpful, thanks!

[@anushkagupta1489]

Hey I am working on the same project and I have some problem while training the model. So can you help me out. Can you please send me your number so that I can ask my doubt as I have my hackathon tomorrow and its really urgent

[@user-lk3zh1wg6y]

Great 🩵🙌🏻

[@kuleenthejas8382]

can i know the source from which u downloaded the image dataset other than the one u mentioned in the video cause i want to train the model to detect Ash trees ,any suggestions would be of great help thank u

[@pranaythapak7079]

Great work!!

[@ComputerVisionEngineer]

Thank you! 😃

[@mksmirnoff3159]

This tutorial was very helpful for me. Thank you! Is it possible to learn more about visual relationships?

[@ComputerVisionEngineer]

😊 I am glad you found it helpful! Training a model with visual relationships type of annotations would be very cool! I will try to do it. 🙌

[@tello9504]

This is a great tutorial!!! the problem is that I don't see exactly how do you test it with the videos. I can't follow that part of the tutorial

[@Panda_2397]

Thank you for your video. It is absolutely amazing that we can have such a great tutorial. Would you please make a detailed tutorial on how to use the model to test real data of video.

[@ComputerVisionEngineer]

Thank you! If you want to inference frames from a video please take a look at the script predict_video.py, it is in the project's github repository. 🙌

[@Panda_2397]

@@ComputerVisionEngineer May I ask how many times you train and the epochs for training to get the model detecting object, I tried to implement the model but it fails to detect. The confusion matrix shows the prediction always wrong😂

[@hbrt10]

hi, thank you for the documentation. I have a problem about predict images. i trained my model and predict image grayscale but i come into view error : ValueError: axes don't match array. What should I do? I must predict image grayscale.

[@castle_of_jokes]

TYSMMM this absolutely helped me create my first object detection model TYSMMMMMMM

[@ComputerVisionEngineer]

You are welcome! 😃 Glad you find it useful! 🙌

[@castle_of_jokes]

@@ComputerVisionEngineer yep defiantly the best course abt the topic on youtube

[@castle_of_jokes]

@@ComputerVisionEngineer are you considering doing a yolo nas course??

[@TheAdrie1987]

thx u sir!

[@mohammadraselamin1473]

your video looks good, I will try it for traffic light detection in complex bg on own custom dataset. if possible, please provide a guide line. you can also make a tutorial , which will be cover lot of topics

[@ComputerVisionEngineer]

Hey Mohammad, well you could use this video as guideline, once you have collected and annotated the data the process is as described in the video. Sounds like a cool project by the way! 😃💪

[@alexfilip1729]

Amazing Tutorial :) I've been looking for this exact video for weeks. Big Big Ups. A small question though, how can you extract the IoU and all the other performance metrics (which appear in the train folder as graphs) in a numerical number (like in a .csv or .txt file)?

[@ComputerVisionEngineer]

Hey Alex, thank you! I am glad you enjoyed it. 😊 In the train directory you have a csv file called results.csv, there is some info there. In order to save the data for the remaining plots (P_curve.png, F1_curve.png, etc.) I suggest you to take a look into ultralytics.ultralytics.yolo.utils.metrics.py. 💪🙌

[@samarasinghesupunc]

Hi I have a question. How we should add the images and txt files on images & labels directories if we have more that 1 class. ex: classes = 2, create 2 folders of classes in images folder then add images?

[@reyniercastano625]

Thanks a lot

[@Reincarnated_Recap]

very helpfull!

[@desireexavier1009]

o senhor é o cara mano

[@yannis_dla]

how would i use yolov8 to count the number of times the object appears in a video? im trying to make a code which counts droplets from an IV infusion. thank you!

[@Falconoo7383]

Thank you very much for the video. After following these videos, I can successfully run this command, but I have an extra few images that I want to detect. How can i use this model to detect?

[@nerpagarpushkar2164]

i didint get the labeling part as we didnt use anywhere can you please guide for that as it would help for more personalised complex detection

[@juangomez2014]

crack excelente tutoríal, pero tengo un problema. Estoy usando una MacBook Air m1, cuando entreno el modelo con la CPU todo bien, pero cuando empleo el MTS para usar la GPU entrena normal pero no es capaz de predecir nada. Que podría ser?? Inclusive cuando miro las gráficas de resultados del modelo entrenado con GPU no sale nada. Gracias

[@frankdearr2772]

Great, Thanks :) 👍

[@hamzahamdi568]

Thank you for all your efforts, i wanna ask you can you give us a roadmap to learn machine learning ? Thank you another time

[@ComputerVisionEngineer]

Hey, sure, I will be happy to make a machine learning roadmap! 😃💪

[@netq254]

Thank you for the amazing help, I just have one question involving the car and license plate detection model you used in your video "Automatic number plate recognition with Python..." I am using the same data set you used in that video, but in the dataset directory is "test, train" and "valid", does this mean I still have to annotate the images myself or has that step already been completed? Thank you!

[@mp051998]

If you downloaded it in the yolo format and setup the directory structure as required, it should work fine. If you got the annotated images from openimages, then check if you have the annotation files (text files with the same name as the images). The directory structure would be something like this if I am not wrong: ├── data/ │ ├── images/ │ │ ├── train/ # Training images (image_name.jpg) │ │ ├── test/ # Testing images (image_name.jpg) │ │ └── val/ # Validation images (image_name.jpg) │ │ │ ├── labels/ │ │ ├── train/ # Training image annotations (image_name.txt) │ │ ├── test/ # Testing images (image_name.txt) │ │ └── val/ # Validation images (image_name.txt) Also make sure the yaml file corresponds to the correct directories. Hope this helps!

[@joannaabuzek3146]

Hi again, I execute all of these things and at the end I got this message: Plotting labels to runs/detect/train3/labels.jpg, but in this directory exists one file named args.yaml and directory named weights, which is empty.... what went wrong? Thanks in advance for your answer....

[@josemariadepereda6237]

best tutorial i found about training with volov8. congrats. On the final stage, you do some prediction tests with videos. Could you guide usto do that test but with images uploaded by the user? Example: user upload an image (photo), and after its uploaded the system detects objects on it. Thanks in advance! you have my suscription and thumbs up allways!

[@josemariadepereda6237]

never mind. found tutorial on oficial ultralitycs docs

[@ComputerVisionEngineer]

Thank you for your support! Inference is covered in other videos, for example this one is about object detection on frames of a video kzbin.info/www/bejne/oHq1g6h9g5Wda6c 🙌

[@ikramessafi9560]

Thank you a lot, I would like to know how to exploit the results of a trained model on unseen data.

[@alexred9963]

Can I use 3d models for education? and than work with photo recognition? I mean for multiple objects much simple find 3d model than lot of images.

[@rajaaimannukmanrajazaharih8960]

hello sir i have another question, how do i extract the coordinates of the bounding box so i can draw a centroid of the bounding box?