๐ CSP-Darknet-53 Mish model
This model is pretrained on ImageNette. The CSP-Darknet-53 Mish architecture was introduced in this paper. It is designed for image classification tasks.
๐ Quick Start
This CSP-Darknet-53 Mish model is pretrained on ImageNette and can be used for image classification tasks. You can follow the installation and usage instructions below to get started.
โจ Features
- Pretrained on ImageNette.
- The architecture adds cross stage partial blocks in the convolutional stage and replaces activations with Mish.
๐ฆ Installation
Prerequisites
Python 3.6 (or higher) and pip/conda are required to install Holocron.
Latest stable release
You can install the last stable release of the package using pypi as follows:
pip install pylocron
or using conda:
conda install -c frgfm pylocron
Developer mode
Alternatively, if you wish to use the latest features of the project that haven't made their way to a release yet, you can install the package from source (install Git first):
git clone https://github.com/frgfm/Holocron.git
pip install -e Holocron/.
๐ป Usage Examples
Basic Usage
from PIL import Image
from torchvision.transforms import Compose, ConvertImageDtype, Normalize, PILToTensor, Resize
from torchvision.transforms.functional import InterpolationMode
from holocron.models import model_from_hf_hub
model = model_from_hf_hub("frgfm/cspdarknet53_mish").eval()
img = Image.open(path_to_an_image).convert("RGB")
config = model.default_cfg
transform = Compose([
Resize(config['input_shape'][1:], interpolation=InterpolationMode.BILINEAR),
PILToTensor(),
ConvertImageDtype(torch.float32),
Normalize(config['mean'], config['std'])
])
input_tensor = transform(img).unsqueeze(0)
with torch.inference_mode():
output = model(input_tensor)
probs = output.squeeze(0).softmax(dim=0)
๐ Documentation
The core idea of the author is to change the convolutional stage by adding cross stage partial blocks in the architecture and replace activations with Mish.
๐ License
This project is licensed under the Apache-2.0 license.
๐ Citation
Original paper
@article{DBLP:journals/corr/abs-1911-11929,
author = {Chien{-}Yao Wang and
Hong{-}Yuan Mark Liao and
I{-}Hau Yeh and
Yueh{-}Hua Wu and
Ping{-}Yang Chen and
Jun{-}Wei Hsieh},
title = {CSPNet: {A} New Backbone that can Enhance Learning Capability of {CNN}},
journal = {CoRR},
volume = {abs/1911.11929},
year = {2019},
url = {http://arxiv.org/abs/1911.11929},
eprinttype = {arXiv},
eprint = {1911.11929},
timestamp = {Tue, 03 Dec 2019 20:41:07 +0100},
biburl = {https://dblp.org/rec/journals/corr/abs-1911-11929.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
Source of this implementation
@software{Fernandez_Holocron_2020,
author = {Fernandez, Franรงois-Guillaume},
month = {5},
title = {{Holocron}},
url = {https://github.com/frgfm/Holocron},
year = {2020}
}