An implementation of SENet, proposed in Squeeze-and-Excitation Networks by Jie Hu, Li Shen and Gang Sun, who are the winners of ILSVRC 2017 classification competition.
Now SE-ResNet (18, 34, 50, 101, 152) and SE-Inception-v3 are implemented.
python exec.py
runs SE-ResNet50 with Cifar10 dataset.
For SE-Inception-v3, the input size is required to be 299x299 as original Inception.