Instance segmentation YOLO v8 | Opencv with Python tutorial

The new version of YOLO v8 by Ultralitycs has recently been released and thanks to its flexibility of use it allows you to classify, detect, and segment objects.
In today's exercise, I'm going to show you how to do object segmentation the easy way.
Blog: pysource.com/2023/02/21/insta...
➤ Courses:
Full Computer Vision course: pysource.com/object-detection...
Training Mask R-CNN PRO (Notebook + Mini-Course): pysource.com/mask-rcnn-traini...
Blueprint: pysource.com/blueprint
➤ Follow me on:
LinkedIn: / pysource
➤ For business inquiries:
pysource.com/contact
#YOLOv8 #segmentation #deeplearning

[@stevenotarealnamematthews140]

FYI I had to change file yolo_segmentation.py line 18 from "result.masks.segments" to "result.masks.xyn" to make it work. Seems like there are some different names depending on which ultralytics package you have installed. Hope that helps.

[@joshteixeira6750]

Another great one!

[@anurag4794]

Amazing!

[@RoboSidekick]

thank you for this video. we are also working on depth mapping with yolo v8 and Open CV.

[@GAMBA-]

Awesome, thanks

[@lushankshambharkar124]

Your tutorial are so easy to understand. Thanks for such tutorials. Thank you Sergio Canu🤗

[@lushankshambharkar124]

I have few questions, can we create our own class ? and can you upload it for the segmentation in videos also

[@roboticverse]

True

[@dlep9221]

Nice ! Thanks

[@emmanuelsheshi961]

thank you so much

[@edytrihatmoko]

Hi, thank you for sharing the video! How to put color inside the ball instead of the borderline, as shown in the video? In case I don't want to show people about the type of it. Thanks!

[@flaviorodrigues5]

Excellent tutorial. I'm trying to create an automatic detection of several species of fish in a static image (a box with several fish). The final idea is to automatically detect different species, classify them and also do an automatically measure of the fish. What videos do you recommend to try to create something similar? Thanks!

[@neverninetofive]

Thank you very much for this video, it’s really helpful! Could you maybe do a tutorial on how to train yolov8 with custom data?

[@RoboSidekick]

we will have one soon on our channel.

[@sarahnawoya498]

Awesome work, Thank you. I would like to create object masks after the segmentation. Is it possible?

[@zhoubin5831]

it is very useful for me. and i also trained yolov8 with my own data, all some round droplets, but after I segment it, its polygon had a lot of very obvious straight lines and not a circle. can yolo fix this problem?

[@saqibqamar9270]

Thank you for your video, it is very helpful. Could you make about how can only get masks in the image?

[@perryfisch2003]

Hi, thank you for taking the time to do this video! I'm constantly amazed at how "simple" things are getting in terms of image processing. I was wondering if YOLO* could be used to identify unknown objects. As you well know YOLO has a defined list of objects that the model was trained for. Could YOLO be adapted/trained for non-modeled objects? I'm working on a robot platform with a forward facing cam pointed downward at a 30 degree angle. There's no fixed list of objects the robot will encounter blocking its way, so I can't train for everything... My goal is "simple" I'm going to move the robot forward and try to navigate around objects that are in the robot's path. YOLO may be overkill here, but I'm trying to learn OpenCV it seemed like a good start. Any alternative suggestions/examples would be more than appreciated! Thank you again for making a complex topic easy to understand.

[@ahmadkhattab4830]

Your tutorials are very nice. Could you make a video for an object tracking with andriod camera if it is possible?

[@md.sabbirrahmanakash7083]

What is the best model to identify and find the area of defects of leather?

[@abdurrahmana.7121]

How can i train custom dataset based segmentation model on yolov8. Please add a tutorial on colab.

[@sasanmir8122]

can someone help me out i want use yolo v8 seg for lane detection and i don't know how to get the center polyline between the two polylines

[@Yonglee7015]

Can you crop the ball based on polygon?

[@russellseangonzalve4566]

Does this work in Yolov5 model?

[@musthaqahmed4497]

Hi, where is the link to download the yolo_segmentation file?

[@raghavsharma4398]

How can i configure my GPU to run this code. I have GTX 1650.

[@GrigoriKog]

How about an interface so other people can see the results.Is that possible?

[@edgarguzman2709]

Can you do a tutorial on how to train the YoloV8 algorithm?

[@raynansouza4262]

What version of python are you using? I had trouble with 3.11. I think pytorch doenst work on it

[@pysource-com]

I used python 3.9

[@raynansouza4262]

@@pysource-com thank you. I am installing this version to try again. :)

[@nhatminhduong1741]

i got an issue it said bus error(core dumped) after i had run the python file. Can someone help me please.

[@mikhaildenisov5307]

If I use YoloV8 do I need to pay money ?

[@shantilalzanwar8687]

This code gives error and not finding solution. Does anyone faced similar issue before ?

[@lazerusmfh]

Do you freelance work? Can I talk to you about some work with yolo and image processing? EDIT: I Emailed you.

[@shantilalzanwar8687]

This code is giving below error : bboxes, classes, segmentations, scores = ys.detect(img) ^^^^^^^^^^^^^^ File "C:\personal\AI ML\ImageSeg001\yolo_segmentation.py", line 18, in detect for seg in result.masks.segments: ^^^^^^^^^^^^^^^^^^^^^ AttributeError: 'NoneType' object has no attribute 'segments' 0: 448x640 1 sports ball, 494.7ms Speed: 3.0ms preprocess, 494.7ms inference, 1.0ms postprocess per image at shape (1, 3, 448, 640) Process finished with exit code 1

[@shantilalzanwar8687]

Thanks for replying, now error is Error. nthreads cannot be larger than environment variable "NUMEXPR_MAX_THREADS" (8) 0: 448x640 1 frisbee, 370.0ms Speed: 3.0ms preprocess, 370.0ms inference, 3.0ms postprocess per image at shape (1, 3, 448, 640) @@Asat48

[@stevenotarealnamematthews140]

Try changing "result.masks.segments" to "result.masks.xyn". Does that help?