Specify the class for detection on pre-trained YOLOv8

[turinaf]

Search before asking

• I have searched the YOLOv8 issues and discussions and found no similar questions.

Question

I wanted to use the yolov8 pre-trained model to detect only certain classes, not all the classes on which the model is trained.
I know we can do this on yolov5 to specify the class. I couldn't do the same with yolov8.

model = torch.hub.load("ultralytics/yolov5", 'custom', path='yolov5s.pt')
model.classes = [0]  # Only person
model.conf = 0.5

Is there something similar for YOLOv8?

Additional

No response

[Laughing-q]

@turinaf

from ultralytics import YOLO
model = YOLO(f"yolov8n.pt")
model.predict(source="ultralytics/assets/", save=True, classes=[0,1])

[turinaf]

@Laughing-q hi from Chengdu,
Thank you!

[kosongdansatu]

@turinaf

from ultralytics import YOLO
model = YOLO(f"yolov8n.pt")
model.predict(source="ultralytics/assets/", save=True, classes=[0,1])

where this code replace on yolov8?

[kosongdansatu]

I'm Running this code from vscode how to select Specify class?

[glenn-jocher]

@kosongdansatu to select specific classes while using YOLOv8 in Visual Studio Code, you can add the "classes" parameter to the predict function, and set it to be a list of integers representing the classes you want to detect. For example, if you only want to detect objects from classes 0 and 1, you can set classes=[0, 1].

Here's an example code snippet:

from ultralytics import YOLO

model = YOLO("yolov8.pt")

# Detect objects from classes 0 and 1 only
classes = [0, 1]

# Set the confidence threshold
conf_thresh = 0.5

# Set the source of the input data (e.g., image file, video file, or folder containing images)
source = "path/to/your/data"

# Call the predict function with the specified parameters
results = model.predict(source=source, save=True, classes=classes, conf=conf_thresh)

Please note that you need to replace "path/to/your/data" with the actual path to your data. Additionally, you can adjust the value of the "conf_thresh" parameter to modify the sensitivity of the model.

[kosongdansatu]

thank you soo much glenn

[glenn-jocher]

@kosongdansatu you're welcome! If you have any further questions or concerns about YOLOv8, don't hesitate to ask. We're always here to help.