I have a small dataset of about 220 images for three classes. I am using YOLO (you only look once) network for an object detection. I am trying to use Active learning in order to reduce the number of training images. I am following this paper.
I trained a baseline model with around 140 images and then used this initial model to select a batch_size (32) from the rest of the data, based on the high uncertain images (low confidence).
I reached a good mean average precision with a low number of samples, but I am getting fluctuations from a round to another, and drop in the last round, that I cannot explain.
The network has the same hyper-parameters, only the data augmentation used is random in each round. I wonder if it is about the size of the batch, or the augmentation or the number of samples for each class from each selection, any hints on how to fix this?

enter image description here


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.