This is a experiment of OctConv with ResNet-50 on CIFAR-10/CIFAR-100
Y. Chen, H. Fang, B. Xu, Z. Yan, Y. Kalantidis, M. Rohrbach, S. Yan, J. Feng. Drop an Octave: Reducing Spatial Redundancy in Convolutional Neural Networks with Octave Convolution. (2019). https://arxiv.org/abs/1904.05049
1. Tensorflow==1.13.1
2. tqdm
Training network on CIFAR-10/CIFAR-100
python train.py --problem cifar10/cifar100
Training the normal ResNet-50
python train.py --is_octconv=false
Training the OctConv ResNet-50
python train.py --is_octconv=true
Training loss
Test accuracy
| Parameters | |
|---|---|
| Normal Resnet | 38,103,690(100%) |
| OctConv-Resnet | 24,628,068(64.5%) |
| alpha | Test accuracy | |
|---|---|---|
| Normal Resnet | 0 | 84.11% |
| OctConv-Resnet | 0.125 | 85.36% |
| OctConv-Resnet | 0.25 | 84.83% |
| OctConv-Resnet | 0.5 | 82.48% |
1. https://github.com/koshian2/OctConv-TFKeras
2. https://github.com/taki0112/ResNet-Tensorflow