the purpose of photometric augmentation #19
Description
Hello, I'm a freshman and I've been reading this article recently, I'd like to ask what is the purpose of photometric augmentation in the article, is it just to increase the dataset, and why can it be effective in the target domain after photometric augmentation, is it because photometric augmentation has a greater effect on the background and a lesser effect on the crowd, thus weakening the effect of the background information on the whole crowd image?
Because I found a paper in the references of this paper that also uses the augmentation technique for domain generalization, which says: ‘To achieve high performance even in unseen domains with unknown sparsity, we focus on one of the main domain gaps in LiDAR data, sparsity difference, and augment domains with diverse sparsity. To this end, scan lines in the source domain are randomly dropped in stead of individual points to simulate the local structures of real LiDAR data .’
So I wanted to understand why this paper chose to perform photometric augmentation.