Hi guys, I just want an advice on the use of yolov8 for my project.
I have my custom dataset, and I want to use yolov8 as detector only, and then use a classifier such as EfficientNet or someother, to classify what yolo detects. what I want to do is:
- take the pre-trained yolov8 model on the coco dataset
- retrain it on my custom dataset using the single_cls = true variable so that I ignore the classification task (?)
- take a classifier and train it on my custom dataset
- pass the regions detected by yolo to the classifier and see the results.
is this a good plan? Is this the correct way to use YOLOv8?
P.S. I have a doubt about the dataset annotation. If there is the object I need to annotate on the ground for like 10 seconds in the same exact position, but around it there are other objects in movement that I need to annotate, should I annotate that object for every frame or should I annotate it only once?
example: a car stopped at the trafficlight for 10 seconds. In these 10 seconds there are people walking. So for all the frames I annotate all the people, but should I annotate the car too despite it is the same identical bounding box?