search

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
http://pjreddie.com/darknet/
Other
21.74k stars 7.96k forks source link

About label? #5682

Open IDGate-ai opened 4 years ago

IDGate-ai commented 4 years ago

If there some objects (ROI) in an image without labels, will this harm the performance? i.e. Do yolo use negative mining?

IDGate-ai commented 4 years ago

Sorry! @AlexeyAB, I found that your paper that answer my question. In Bag of freebies section, "....In dealing with the problem of semantic distribution bias, a very important issue is that there is a problem of data imbalance between different classes, and this problem is often solved by hard negative example mining [72]or online hard example mining [67] in two-stage object detector. But the example mining method is not applicable to one-stage object detector, because this kind of detector belongs to the dense prediction architecture. Therefore Lin et al. [45] proposed focal loss to deal with the problem of data imbalance existing between various classes."

AlexeyAB commented 4 years ago

https://github.com/AlexeyAB/darknet#how-to-improve-object-detection

check that each object that you want to detect is mandatory labeled in your dataset - no one object in your data set should not be without label. In the most training issues - there are wrong labels in your dataset (got labels by using some conversion script, marked with a third-party tool, ...). Always check your dataset by using: https://github.com/AlexeyAB/Yolo_mark

desirable that your training dataset include images with non-labeled objects that you do not want to detect - negative samples without bounded box (empty .txt files) - use as many images of negative samples as there are images with objects

HYL-Dave commented 4 years ago

check that each object that you want to detect is mandatory labeled in your dataset - no one object in your data set should not be without label. In the most training issues - there are wrong labels in your dataset (got labels by using some conversion script, marked with a third-party tool, ...). Always check your dataset by using: https://github.com/AlexeyAB/Yolo_mark

desirable that your training dataset include images with non-labeled objects that you do not want to detect - negative samples without bounded box (empty .txt files) - use as many images of negative samples as there are images with objects

  1. If I want to classify 3 class objects with 10000 images, then are there 10000 negative samples (empty .txt files)?
  2. Suppose that I want detect dogs. Now there an image with two dogs. But there is only one label in this image (correct position). Does this decrease the accuracy due to miss a label?
AlexeyAB commented 4 years ago
  1. I don't know are there or aren't, but there should be )
  2. If you missed some labels - it will decrease accuracy