Open rangganast opened 11 months ago
Normalization takes place by the framework(s) automatically - it's done at runtime, including resizing with/without considering aspect-ratio when (re-)training the model: the train/val images have a resolution of 480x270, but the training happens with imgsz=640
.
Greetings,
I recently studied about Object Segmentation Model and I got some guides thanks to this linkhttps://learnopencv.com/train-yolov8-instance-segmentation/. However, I still really don't know how to normalize the trashcan dataset. I downloaded the original of the trashcan dataset and it's still not normalized (point value not between 0 and 1). Since YOLOv8 requires the user to use the normalized labels dataset.
My question is how to normalize the object segmentation dataset like you did on the trashcan dataset? I need it to apply on other datasets also for my practice.
Thank you so much.