YOLOv8 COMPLETE Tutorial | Object Detection | Segmentation | Classification

[@andrespereira4852]

Hi buddy. Always coming with an concise and clear video about all versions, always making a excellent videos. I am now running Yolov8 on my computer in less than 5 minutes. The only drawback is that after running the terminal command, yolo doesn't save the predictions at all. Have you any clue why is that? In fact, in terminal I dont get the line: "Results saved to..."

[@andrespereira4852]

😅😅 ultralytics now use flag "save=True" to save results.

[@TheCodingBug]

Yes. save=True has to be set for saving output.

[@charse09]

Was just wondering the same thing. Thanks for asking!

[@debanjanchakraborty9946]

I was working with yolov7 but now that the version has changed now I'll take this best explanation as a base for my future projects

[@TheCodingBug]

To my experience, yes it is better for smaller objects in the frames.

[@HasanKarakus]

this is not just a video, this is art.

[@brunospfc8511]

Thanks, will be trying this morning, great video!

[@juanzmi3700]

Simple, straight to the point. To many more videos from you!

[@Ddangcho-i8g]

Your lecture is innovative. Thanks to you, I could learn a lot. Thank you again.

[@cristipanaintescu605]

had some derby while installing torch and ultralytics and with the paths but in the end I really got so far and learnt so much, so in conlcusion, muchas gracias aficion, esta es para vosotros SIUUU

[@paulpolizzi3421]

Wow Amazing !!! I wouldn't know where to start , but would love to know if this can work on my kids soccer games, taking touches of the ball , passing? goals? corners? possession percentage? would love to learn or could i send you a video to try?

[@shubhamgade1375]

I am going to switch from detection2 to YOLO v8. I am desperately waiting for custom object segmentation using YOLOv8. It’s fantastic. Thanks for the video.👌🏻👍🏻

[@TheCodingBug]

The tutorial for custom object segmentation will be out on Monday next week.

[@ardumaniak]

​@@TheCodingBugI'll be waiting!

[@kishorem5530]

Big follower of your presentation brother...Thank you

[@Ramp_cat_7]

this is exactly what i need thanx

[@christian.js.1997]

I'm using YOLOv7 for our research, how can I count the total detected instances or objects in a video?

[@TheCodingBug]

You can initialize an empty dictionary and increase count against each class label for each detection. Then print that dictionary to get counts against each class label. This should be in the loop where you plot a bounding box for each class.

[@GuilhermePiresSilvadeAlmeida]

Congratulations on the video. My notebook doesn't have a dedicated video card, so I choose CPA instead of CUDA. Correct? Wouldn't it be possible to leave a link to the image files used in the example in the description? Thank you very much in advance.

[@TheCodingBug]

You should skip installing pytorch GPU. Rest of the things remain the same and the model will use CPU automatically. You can take images from Google.

[@sumitsingh4245]

hi, how to make this without command prompt, can we make this with VS code or Jupiter notebook

[@lucianoortenzi]

I really appreciate your videos! Very usefull and instructive! When will you come out with object counting with YOLOv8? :D

[@TheCodingBug]

Custom object detection first and then perhaps objects tracking and finally object counting.

[@Kalkiigm]

Can you make a video on YOLOv8 classification specifically after postprocessing the imagenet values?

[@alfibima4247]

I'm trying YOLOv8 to detect live video streaming on KZbin. but the speed is very slow. how to boost the speed so that object detection can be realtime? I use GTX 1650 GPU

[@itaynave5770]

Thanks, great video! How can I extract the bounding boxes in python? (dont save it to txt file)

[@EgeSener]

when i open a python file i can't import YOLO from ultralytics. it says it isnt defined but it works on terminal. Can you help me ?

[@benmaslen7041]

Like the video! I have cloned the yolov8 github repo, installed the packages in my python virtual environment, however the command line does not recognise the 'yolo' command? Any tips? "yolo: command not found"

[@patrickruci]

Best Tutorial out of every single on I've watched, though I have a little question. How do I find the folder where I should put all my pictures/videos in? I already tried it in the normal yolov8 folder, it doesn't work. When I have the pictures in a folder on my desktop, it says that the image doesn't even exist. It's all set up in: "anaconda3\envs\yolov8"

[@TheCodingBug]

You need to give absolute path of the folder like C:\Users\Desktop\Folder

[@patrickruci]

​@@TheCodingBug (yolov8) C:\Users\Nutzer>yolo task=detect mode=predict model=yolov8n.pt source=C:\Users\Nutzer\Desktop\YOLO\peopleonstreet.jpg When I try that it says that theres a backend fallback

[@josecarlosramirez2682]

Is there a way to detect a specific object, for example only people omitting cars, buses, etc. ?

[@hussain9149]

well and easily explained

[@firasnizam]

thanks for sharing useful information

[@MrtaiDiary]

thank you for your video . could you please yolov8 custom data best.pt to use in android app kotlin ?

[@kennychen7606]

nice video! I’m very interested in image segmentation!

[@itayhilel2168]

How dose the rspt work? Can you show an example??

[@thewajahatimam]

Can we use this while runtime detection like with camera

[@TheCodingBug]

Yes. I've mentioned in the video how to run object detection on webcam.

[@io.15-il4jt]

thanks bro

[@christopherfranceschi5886]

how would i change the size of the preview window. (inf frames of shape 640x480 at 30.00 FPS) --< I want to make it larger

[@TheCodingBug]

Resize the frame that is being displayed to a new resolution of your choice....

[@zerosploit]

great video

[@alfibima4247]

How to optimize yolov8 speed with tensorrt?

[@dataScineceEnthusiast]

I am getting this 'yolo' is not recognized as an internal or external command, operable program or batch file. pls help

[@TheCodingBug]

Install ultralytics after running the command prompt as administrator and also make sure that your environment is activated in which you've installed ultralytics. Or Use Python script instead of command line.

[@dataScineceEnthusiast]

@@TheCodingBug i checked yolov8 is showing in the conda env (navigator) and i have installed ultralytics everything as you described

[@swarupdas4321]

@@dataScineceEnthusiast same error

[@dataScineceEnthusiast]

@@swarupdas4321 I tried Python Script and it worked for me. You can also try that.

[@ssarswat1983]

@@dataScineceEnthusiast Can you kindly upload a screenshot please. Facing same error.

[@vipulanb5581]

how to extract mask and consider as a Contour for area extraction

[@TheCodingBug]

save_txt gives you mask polygon which you can convert into segmentation mask

[@immaculate6244]

To hide boxes you can just pass in the argument: boxes= False

[@razorphone77]

I noticed that when i fed in a video file made of images that were 250x250, it looks like its converted them to 640x640 and you mention that in the video, that its trained on 640x640. Can you stop it upscaling the input? because I think its causing issues in the inference speed. ie taking far too long. I really need to feed in a video steam of about 30fps at 250x250. Does using the cpu version of torch also limit the inference speed? Or does it just impact training time? because I ended up with maybe 2-3fps on inference using Yolo(medium) and I know it should be able to do real time video, so what gives? In fact your video of the cars works fine....?

[@TheCodingBug]

I always use GPU. CPU would always give you 3-4 FPS. Use nano version with GPU for real time FPS.

[@razorphone77]

@@TheCodingBug but I know that mobile phones can run the kind of traffic video you showed. They don't have GPUs

[@vimalkumar2220]

RuntimeError: DataLoader worker (pid(s) 3580, 4732, 13692, 1840, 9632, 9648, 14032) exited unexpectedly Iam getting this error while train anyone help pls

[@3bones383]

build a machine learning model that can detect various objects in images. Test and train.csv help me 😓

[@namquach73]

I don't get image detection output

[@TheCodingBug]

Set save=True show=True

[@zultandimitry572]

We need comparison on a same machine between the v8 v7 and v5 to see the real difference

[@AuxiliaryPanther]

Yeah. Please post the video when you finish the comparison.

[@alinanaz2983]

How to print FPS ?

[@TheCodingBug]

You'll need to modify predict file of ultralytics library. Or clone ultralytics repository and modify predict.py and run yolo directly from that repository instead of installing it with pip.

[@SivaBodapati]

How to print Accuracy of detected object 🙄