Readspeaker menu

CNN Models

We provide convolutional neural network models which were trained to classify the 1000 classes of the ImageNet LSVRC 2012 - 2016 classification challenge. This includes:

  • AlexNet and VGG16 with batch normalization added
  • Residual Networks

The pre-trained models are available at https://github.com/cvjena/cnn-models/

Results

Single-crop validation error rates on ILSVRC 2012 - 2016.

Model Top-1 error (vs. original)
Top-5 error (vs. original)
AlexNet_cvgj  39.9% (vs. 42.6%)  18.1% (vs. 19.6%)
VGG19_cvgj  26.9% (vs. 28.7%)  8.8% (vs. 9.9%)
ResNet10_cvgj  36.1%  14.8%
ResNet50_cvgj  24.6% (vs. 24.7%)  7.6% (vs. 7.8%)