This repo contains pre-trained models by Dense-Sparse-Dense(DSD) training on Imagenet.
Compared to conventional training method, dense→sparse→dense (DSD) training yielded higher accuracy with same model architecture.
Sparsity is a powerful form of regularization. Our intuition is that, once the network arrives at a local minimum given the sparsity constraint, relaxing the constraint gives the network more freedom to escape the saddle point and arrive at a higher-accuracy local minimum.
Feel free to use the better-accuracy DSD models to help your research.
| Baseline | Top-1 error | Top-5 error | DSD | Top-1 error | Top-5 error |
|---|---|---|---|---|---|
| AlexNet | 42.78% | 19.73% | AlexNet_DSD | 41.48% | 18.71% |
| VGG16 | 31.50% | 11.32% | VGG16_DSD | 27.19% | 8.67% |
| GoogleNet | 31.14% | 10.96% | GoogleNet_DSD | 30.02% | 10.34% |
| SqueezeNet | 42.39% | 19.32% | SqueezeNet_DSD | 38.24% | 16.53% |
| ResNet18 | 30.43% | 10.76% | ResNet18_DSD | 29.17% | 10.13% |
| ResNet50 | 24.01% | 7.02% | ResNet50_DSD | 22.89% | 6.47% |
The beseline of AlexNet, VGG16, GoogleNet, SqueezeNet are from Caffe Model Zoo. The baseline of ResNet18, ResNet50 are from fb.resnet.torch commit 500b698.