No project description provided
Project description
Resnets
Tidy implementation of classical neural networks for classification.
Available frameworks:
- Jax
- PyTorch
- TensorFlow
Everything in one place with results matching those reported in papers.
Installation
Install the package with pip:
pip install resnets
Available architectures
VGG
architecture | parameters | reported best | this repository | CIFAR-10 ckpt | ImageNet ckpt |
---|---|---|---|---|---|
VGG11 | 9.2M | 7.81 | 7.98 | ||
VGG13 | 9.4M | 6.35 | 6.17 | ||
VGG16 | 14.7M | 6.49 | 6.34 | ||
VGG19 | 20.0M | 6.76 | 6.72 |
Sources:
- Very Deep Convolutional Network for Large-Scale Image Recognition
- On Correlation of Features Extracted by Deep Neural Networks
ResNet
architecture | parameters | reported best | this repository | CIFAR-10 ckpt | ImageNet ckpt |
---|---|---|---|---|---|
ResNet-20 | 0.27M | 8.75 | 7.99 | ||
ResNet-32 | 0.46M | 7.51 | 7.40 | ||
ResNet-44 | 0.66M | 7.17 | 6.83 | ||
ResNet-56 | 0.85M | 6.97 | 6.23 | ||
ResNet-110 | 1.7M | 6.37 | 5.98 | ||
ResNet-164 | 1.7M | 5.46 | 5.27 | ||
ResNet-1001 | 10.3M | 4.92 | 5.06 |
Sources:
Wide ResNet
architecture | parameters | reported mean | this repository | CIFAR-10 ckpt | ImageNet ckpt |
---|---|---|---|---|---|
WRN-16-4 | 2.7M | 5.02 | |||
WRN-40-4 | 8.9M | 4.53 | 4.46 | ||
WRN-16-8 | 11.0M | 4.27 | |||
WRN-28-10 | 36.5M | 4.00 | |||
WRN-28-10+ | 36.5M | 3.89 |
(+ with dropout)
Sources:
ResNeXt
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
resnets-0.0.1.tar.gz
(6.0 kB
view hashes)