Open shuntos opened 6 years ago
General rule - your training dataset should include such a set of relative sizes of objects that you want to detect - differing by no more than 2 times:
train_network_width * train_obj_width / train_image_width ~= detection_network_width * detection_obj_width / detection_image_width
train_network_height * train_obj_height / train_image_height ~= detection_network_height * detection_obj_height / detection_image_height
I use yolo v2 , images sizes are different .use BBOX tool for annotation ,converted it into yolo format
Or your dataset is wrong, or the general rule is broken: https://github.com/AlexeyAB/darknet#how-to-improve-object-detection
I run detection on my own custom dataset but it detect object with bigger boundary boxes then its actual dimension ,and unable to detect small object of same class .