This page contains pre-trained image models either converted from other toolkits or trained from scratch with CNTK. The list of available models includes:
CNTK model download path | https://www.cntk.ai/Models/CNTK_Pretrained/AlexNet_ImageNet_CNTK.model |
---|---|
Training script | AlexNet_ImageNet.cntk |
Single crop top 1 / top 5 error | 40.106% / 17.746% |
CNTK model download path | https://www.cntk.ai/Models/Caffe_Converted/AlexNet_ImageNet_Caffe.model (Last updated: April 28, 2017) |
---|---|
Source Caffe model website | https://github.com/BVLC/caffe/tree/master/models/bvlc_alexnet |
Single crop top 5 error | 19.8% |
CNTK model download path | https://www.cntk.ai/Models/CNTK_Pretrained/InceptionV3_ImageNet_CNTK.model (Last updated: August 11, 2017) |
---|---|
Training script | InceptionV3_ImageNet_Distributed.py |
Single crop top 1 / top 5 error | 21.520% / NA |
CNTK model download path | https://www.cntk.ai/Models/Caffe_Converted/BNInception_ImageNet_Caffe.model (Last updated: April 28, 2017) |
---|---|
Source Caffe model website | https://github.com/BVLC/caffe/tree/master/models/bvlc_googlenet |
Single crop top 5 error | 11.50% |
CNTK model download path | https://www.cntk.ai/Models/CNTK_Pretrained/ResNet18_ImageNet_CNTK.model (Last updated: Nov. 16, 2017) |
---|---|
Training script | ResNet18_ImageNet1K.cntk |
Single crop top 1 / top 5 error | 28.752% / 9.700% |
CNTK model download path | https://www.cntk.ai/Models/CNTK_Pretrained/ResNet34_ImageNet_CNTK.model (Last updated: Nov. 16, 2017) |
---|---|
Training script | ResNet34_ImageNet1K.cntk |
Single crop top 1 / top 5 error | 26.114% / 8.386% |
CNTK model download path | https://www.cntk.ai/Models/CNTK_Pretrained/ResNet50_ImageNet_CNTK.model (Last updated: Nov. 16, 2017) |
---|---|
Training script | ResNet50_ImageNet1K.cntk |
Single crop top 1 / top 5 error | 23.358% / 6.740% |
CNTK model download path | https://www.cntk.ai/Models/CNTK_Pretrained/ResNet101_ImageNet_CNTK.model (Last updated: Nov. 16, 2017) |
---|---|
Training script | ResNet101_ImageNet1K.cntk |
Single crop top 1 / top 5 error | 21.822% / 6.042% |
CNTK model download path | https://www.cntk.ai/Models/CNTK_Pretrained/ResNet152_ImageNet_CNTK.model (Last updated: Nov. 16, 2017) |
---|---|
Training script | ResNet152_ImageNet1K.cntk |
Single crop top 1 / top 5 error | 21.300% / 5.760% |
CNTK model download path | https://www.cntk.ai/Models/CNTK_Pretrained/ResNet20_CIFAR10_CNTK.model (Last updated: Nov. 16, 2017) |
---|---|
Training script | TrainResNet_CIFAR10.py --network resnet20 |
Single crop top 1 error | 8.23% |
CNTK model download path | https://www.cntk.ai/Models/CNTK_Pretrained/ResNet110_CIFAR10_CNTK.model (Last updated: Nov. 16, 2017) |
---|---|
Training script | TrainResNet_CIFAR10.py --network resnet110 |
Single crop top 1 error | 6.24% |
CNTK model download path | https://www.cntk.ai/Models/Caffe_Converted/ResNet50_ImageNet_Caffe.model (Last updated: April 28, 2017) |
---|---|
Source Caffe model website | https://github.com/KaimingHe/deep-residual-networks |
Single crop top 5 error | 7.75% |
CNTK model download path | https://www.cntk.ai/Models/Caffe_Converted/ResNet101_ImageNet_Caffe.model (Last updated: April 28, 2017) |
---|---|
Source Caffe model website | https://github.com/KaimingHe/deep-residual-networks |
Single crop top 5 error | 7.12% |
CNTK model download path | https://www.cntk.ai/Models/Caffe_Converted/ResNet152_ImageNet_Caffe.model (Last updated: April 28, 2017) |
---|---|
Source Caffe model website | https://github.com/KaimingHe/deep-residual-networks |
Single crop top 5 error | 6.71% |
CNTK model download path | https://www.cntk.ai/Models/Caffe_Converted/VGG16_ImageNet_Caffe.model (Last updated: April 28, 2017) |
---|---|
Source Caffe model website | http://www.robots.ox.ac.uk/~vgg/research/very_deep/ |
Single crop top 5 error | 10.11% |
CNTK model download path | https://www.cntk.ai/Models/Caffe_Converted/VGG19_ImageNet_Caffe.model (Last updated: April 28, 2017) |
---|---|
Source Caffe model website | http://www.robots.ox.ac.uk/~vgg/research/very_deep/ |
Single crop top 5 error | 10.18% |