Awesome
mxnet-model
MXNet models (only classification for now) and deploy prototxt for resnet, resnext, inception_v3, inception_v4, inception_resnet, wider_resnet, densenet, DPNs and other networks.
We recommend using these MXNet models with official MXNet repository
Please install MXNet for evaluating and finetuning.
Disclaimer
Most of the pre-train models are converted from other projects, the main contribution belongs to the original authors.
Project links:
pytorch-classification、mxnet-model-gallery、 tensorflow slim、craftGBD、 ResNeXt、keras deep-learning-models、 DPNS、Senet
CLS (Classification, more details are in cls)
Performance on imagenet validation.
Top-1/5 error of pre-train models in this repository (Pre-train models download urls).
| Network | 224/299<br/>(single-crop) | 320/395<br/>(single-crop) | Download |
|---|---|---|---|
| resnet101_v2 | 21.89/6.01 | 20.44/5.22 | 170.3MB(BaiduCloud) |
| resnet152_v2 | 20.71/5.42 | 19.65/4.75 | 230.2MB(BaiduCloud) |
| resnet269_v2 | 19.73/4.99 | 18.64/4.33 | 390.4MB(BaiduCloud) |
| resnext50_32x4d | 22.38/6.31 | 21.10/5.52 | 95.8MB(BaiduCloud) |
| resnext101_32x4d | 21.33/5.80 | 19.92/4.97 | 169.1MB(BaiduCloud) |
| resnext101_64x4d | 20.60/5.41 | 19.26/4.63 | 391.2MB(BaiduCloud) |
| inception_v3 | 21.70/5.75 | .. | 91.1MB(BaiduCloud) |
| inception_v4 | 20.03/5.09 | .. | 163.1MB(BaiduCloud) |
| inception_resnet_v2 | 19.86/4.83 | .. | 213.3MB(BaiduCloud) |
| xception | 20.89/5.48 | .. | 87.4MB(BaiduCloud) |
| air101 | 21.32/5.77 | .. | 246.2MB(BaiduCloud) |
| dpn-92-extra | 19.98/5.06 | 19.00/4.37 | 145MB(GoogleDrive) |
| dpn107-extra | 19.75/4.94 | 18.34/4.19 | 333MB(GoogleDrive) |
| se-resnet50 | 22.39/6.37 | 20.50/5.23 | 107.4MB(BaiduCloud) |
| se-resnext50-32x4d | 20.96/5/54 | 19.35/4.66 | 105.4MB(BaiduCloud) |
| se-inception-v2 | 23.64/7.05 | 21.61/5.88 | 45.6MB(BaiduCloud) |
- 224x224(base_size=256) and 320x320(base_size=320) crop size for resnet-v2/resnext, 299x299(base_size=320) and 395x395(base_size=395) crop size for inception.