1 paper
Guangrun Wang, Liang Lin, Rongcong Chen +2
Current state-of-the-art visual recognition systems usually rely on the following pipeline: (a) pretraining a neural network on a large-scale dataset (e.g., ImageNet) and (b) finet…