Metalhead
Metalhead.jl provides standard machine learning vision models for use with Flux.jl. The architectures in this package make use of pure Flux layers, and they represent the best-practices for creating modules like residual blocks, inception blocks, etc. in Flux.
Installation
]add Metalhead
Available models
| Model Name | Function | Pre-trained? |
|---|---|---|
| VGG-11 | VGG11 | N |
| VGG-11 (w/ BN) | VGG11 | N |
| VGG-13 | VGG13 | N |
| VGG-13 (w/ BN) | VGG13 | N |
| VGG-16 | VGG16 | N |
| VGG-16 (w/ BN) | VGG16 | N |
| VGG-19 | VGG19 | N |
| VGG-19 (w/ BN) | VGG19 | N |
| ResNet-18 | ResNet18 | N |
| ResNet-34 | ResNet34 | N |
| ResNet-50 | ResNet50 | N |
| ResNet-101 | ResNet101 | N |
| ResNet-152 | ResNet152 | N |
| GoogLeNet | GoogLeNet | N |
| Inception-v3 | Inception3 | N |
| SqueezeNet | SqueezeNet | N |
| DenseNet-121 | DenseNet121 | N |
| DenseNet-161 | DenseNet161 | N |
| DenseNet-169 | DenseNet169 | N |
| DenseNet-201 | DenseNet201 | N |
Getting Started
You can find the Metalhead.jl getting started guide here: https://fluxml.ai/Metalhead.jl/dev/docs/tutorials/quickstart.html