Viet-Anh on Software Logo

What is: IoU-Balanced Sampling?

SourceLibra R-CNN: Towards Balanced Learning for Object Detection
Year2000
Data SourceCC BY-SA - https://paperswithcode.com

IoU-Balanced Sampling is hard mining method for object detection. Suppose we need to sample NN negative samples from MM corresponding candidates. The selected probability for each sample under random sampling is:

p=NMp = \frac{N}{M}

To raise the selected probability of hard negatives, we evenly split the sampling interval into KK bins according to IoU. NN demanded negative samples are equally distributed to each bin. Then we select samples from them uniformly. Therefore, we get the selected probability under IoU-balanced sampling:

p_k=NK1M_k , k[0,K) p\_{k} = \frac{N}{K}*\frac{1}{M\_{k}}\text{ , } k\in\left[0, K\right)

where M_kM\_{k} is the number of sampling candidates in the corresponding interval denoted by kk. KK is set to 3 by default in our experiments.

The sampled histogram with IoU-balanced sampling is shown by green color in the Figure to the right. The IoU-balanced sampling can guide the distribution of training samples close to the one of hard negatives.