Convolutional Neural Networks
ShuffleNet v2
Introduced by Ma et al. in ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture DesignShuffleNet v2 is a convolutional neural network optimized for a direct metric (speed) rather than indirect metrics like FLOPs. It builds upon ShuffleNet v1, which utilised pointwise group convolutions, bottleneck-like structures, and a channel shuffle operation. Differences are shown in the Figure to the right, including a new channel split operation and moving the channel shuffle operation further down the block.
Source: ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design
Papers
| Paper | Code | Results | Date | Stars |
|---|
Tasks
| Task | Papers | Share |
|---|---|---|
| Image Classification | 4 | 21.05% |
| Object Detection | 4 | 21.05% |
| General Classification | 2 | 10.53% |
| Semantic Segmentation | 2 | 10.53% |
| Adversarial Attack | 1 | 5.26% |
| Adversarial Robustness | 1 | 5.26% |
| Self-Driving Cars | 1 | 5.26% |
| Real-Time Object Detection | 1 | 5.26% |
| Real-Time Semantic Segmentation | 1 | 5.26% |
1x1 Convolution
Convolution
Dense Connections
Global Average Pooling
ShuffleNet V2 Block
ShuffleNet V2 Downsampling Block
Softmax
