CNN Architechture
by Jiutian Yu
1. 1.Deeper Network
1.1. 1.1 VGG16
1.2. 1.2 VGG19
1.3. 1.3 MSRA-Net
2. 2.Reinforce convolution block
2.1. 2.1 NIN (Network in network)
2.2. 2.2 Inception Net v1-v4
3. 4. From classification task to recognition task
3.1. 4.1 YOLO v1/v2
3.2. 4.2 SSD/DSSD
3.3. 4.3 RCNN/SPP-Net/Faster R-CNN/FPN
3.3.1. 4.3.1 Mask R-CNN
4. 5. Multitask
4.1. 5.1 SDS
4.1.1. 5.1.1 DeepMask
4.1.1.1. 5.1.1.1 FCIS
5. 6. Add new functional unit
5.1. 6.1 Inception v2 (BatchNormalization)
5.2. 6.2 FCN/FCN-CRF
5.3. 6.3 STNet
5.4. 6.4 CNN + RNN/LSTM
5.5. 6.5 GAN
6. 3. Intergrated (Deeper, Faster convergence)
6.1. 3.1 ResNet
6.1.1. 3.1.1 ResNext
6.1.2. 3.1.2 DenseNet
6.1.3. 3.1.3 Inception-ResNet
6.1.4. 3.1.4 Xception
6.1.5. 3.1.5 MobileNet/ShuffleNet/SqueezeNet (Light)