I aim to use the YOLO11 model with the Hailo 8L module. I would like to detect people and animals (such as foxes, dogs, and cats, and possibly more in the future). Initially, I’ll work with the standard model and then gradually add custom datasets over time.
What is the difference between the segmentation model and the detection model? From what I understand, both seem to identify the same objects in the frame, but is the main difference that the segmentation model also creates an outline around the object?
Is it also possible to configure the model to identify a generic “animal” instead of specific types (e.g., dog, cat, fox)?
Detection models only produce boxes around the objects. Instance segmentation models produce a mask over the object. Instance segmentation models require more labelling effort because you need to draw the mask on each object as opposed to detection models which just require drawing a box. If you’re not interested in getting a mask, then detection model is easier.
Is it also possible to configure the model to identify a generic “animal” instead of specific types (e.g., dog, cat, fox)?
YOLO11 only can detect what it was trained to detect. So you can’t change what it detects without retraining it on a custom labelled dataset that does what you want.
You can try YOLOE which allows using prompts to specify what you want to detect: