I tried to train my 15k images dataset which includes 14 classes and 5k images with no objects. I search about how to label my no object image and see in the guide of @AlexeyAB said:
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
https://github.com/AlexeyAB/darknet#how-to-improve-object-detection
- However, I don't know where to put the empty.txt file.
- In the case of negative samples, does it mean that I should label my no object as a negative number. For example, in line of annotation txt file, normal class is [14 0.1 0.1 0.1 0.1] but for no object class is just [-1]