Run Segmentation with Pre-Trained Ultralytics YOLOv8 Model in Python | Episode 2

Join Nicolai Nielsen for Episode 2 of our exciting series on leveraging the power of Ultralytics YOLOv8 for object segmentation! In this video, Nicolai delves into the core concepts of object segmentation and showcases the impressive capabilities of the pre-trained YOLOv8 model for running real-time inference. You’ll learn how to set up the YOLOv8 model, run it from the command line, and see it in action with a webcam.

In this tutorial, Nicolai covers:
- An introduction to object segmentation using YOLOv8
- Step-by-step instructions for setting up the model in Python
- Running inference modes directly from the command line
- Real-time segmentation demonstration using a webcam
- Tips on customizing the model for your specific applications

We’ll also explore key documentation to help you tailor YOLOv8 to your projects, ensuring you can harness the full potential of this state-of-the-art computer vision model. Whether you’re an AI enthusiast, a researcher, or a developer, this episode is packed with essential insights to elevate your AI projects.

Don't miss out on the cutting-edge advancements in visual AI—subscribe to our channel, give this video a thumbs up, and check out the links below for more details and resources.

📖 Learn more:

#YOLOv8 #Ultralytics #ObjectSegmentation #AI #ComputerVision #MachineLearning #DeepLearning #RealTimeInference

[agmeister]

comments on here seem weird, it's like people can't read the docs, all while the documentation is insanely vast, may not be perfect, nothing is, but is really darn good.

Anyway, thx a lot for providing such a good framework (i guess we can call it that) and these videos as well, i'm watching the whole series and it's really helping me out with my thesis, never thought it would be so easy to use such advanced models.

[os-EmilyW]

Nicolai, the real-time segmentation demo was quite impressive! I'm curious, how resilient is the YOLOv8 model when it comes to segmenting overlapping or occluded objects, particularly in less-than-ideal lighting conditions? Has anyone tested it in such challenging environments to gauge its robustness?

[LunaStargazer-vs]

Wow, Nicolai, your demo of YOLOv8's real-time segmentation is mesmerizing! I'm curious, how does YOLOv8 handle overlapping objects that are similar in shape and color? Does it maintain accuracy, especially in more chaotic scenes, or does it sometimes get 'confused'?

[AbdullahJirjees]

how to remove the bounding box at the segmention output and just keep the segmentation polygon?

[ollydsouza]

Please give us an idea how we can store the "results" of the detection so it can be stored in a database or excel file?

[AxelRyder-qb]

Yo, Nicolai! Super dope vid!! Quick Q: any tips on optimizing YOLOv8 for lower-end GPUs when doing real-time segmentation with a webcam? 🤔 Might save a bunch of us from dropping $$$ on new hardware! Thanks!!!

[onemilliongecs]

I'm trying to use this model as a parking lot car detection system but it doesn't seem to like any of the sample images or videos I give it. There are a lot of cars in each one so maybe it's unable to handle that many objects on screen at once but it seems to do fine with finding people, chairs, and cell phones through my live webcam feed. Are there any other pre-trained models that are better at finding cars or is there something I'm not changing in my code that can make the default model better at detecting cars?

[arthurmorgan]

Hello, could you drop the code from the end of the video

[aiforyounow]

is it possible to only segment a particular class of object detected

[TravelwithRasel.]

i dunno what's wrong with me, i can not run car detection video, even though when i am going to hit the run button it is showing error of time is not defined, but for the testing i have wrote details as like you. can you tell me the solution of this error

[zaidahmed]

Does YOLO supports 3d image segmentation ? like the files in .obj format

[viktorull]

Hello there, do you still have this code yolov8_segmentation_output.py available? I am trying to run it here (I was typing from your video) but I got an error when I ran it

[RashidKhan-mmrw]

Hello i annotated 200k + Dataset for image segmentation using Yolov8 Please respond to me with the best solutions... Advanced thank you 😉

[AryanSharma-gjff]

Sir for me some file are missing could u send me the link for the above code

[fkkkkdq]

Some files like cv2, numpy, psu pil are missing

[aodexiusi]

Hello here, if I want to change the color of the boxes to black, how can I do it?

[m]

When using the pre-trained YOLOv8 model for segmentation, what are some real-world scenarios where it outperforms traditional methods, and have you encountered any unexpected challenges or limitations?

[AyahuascaDataScientist]

there’s documentation for training models with the MPS framework (Apple m-series chips).

But when you just run inference on, say, your webcam or a video you have in your directory…there isn’t documentation for running it on GPU?

In the video, the guy brags about how he has an NVIDIA GPU, which is why his frame rate is high.

Meanwhile, my M1 Max is churning a 7 FPS on inference, but I’m assuming it’s using the CPU. Is there not documentation for running inference on MPS GPU?

```
from ultralytics import YOLO

# Load a pretrained YOLOv8n model
model = YOLO ('yolov8n-seg.pt')

# Run inference on the source
results = model(source=1, show=True, conf=0.3, save=True)

```

[xxkbkym]

Why you talking so fast?
You trying to save space on YouTube?

[kemaldarc]

My code is like this, but I don't want the object to be enclosed in a square when doing segmentation. How can I change my code for this? Have a good day

"""from ultralytics import YOLO
import matplotlib.pyplot as plt

# Load the segmentation model
model = projesi/yolov8s-seg.pt")

# Run instance segmentation on an image
results = projesi/images.jpg')

# Visualize the segmentation without boxes
annotated_image = results[0].plot() # Remove the 'show_boxes=False' argument

# Show the annotated image
plt.imshow(annotated_image)
plt.axis('off') # optional, turn off axes
plt.show()""""