Densenet121 Res224 Chex

🚀 densenet121-res224-chex

A DenseNet is a type of convolutional neural network. It uses dense connections between layers through Dense Blocks. In these blocks, all layers (with matching feature - map sizes) are directly connected to each other. To maintain the feed - forward nature, each layer gets additional inputs from all previous layers and passes its own feature - maps to all subsequent layers. This model is useful for image classification tasks, especially in the field of vision, specifically for classifying X - ray images.

🚀 Quick Start

How to use

Here is how to use this model to classify an X - ray image:

⚠️ Important Note

Each pretrained model has 18 outputs. The all model has every output trained. However, for the other weights, some targets are not trained and will predict randomly because they do not exist in the training dataset. The only valid outputs are listed in the field {dataset}.pathologies on the dataset that corresponds to the weights.

Benchmarks of the models are here: BENCHMARKS.md

💻 Usage Examples

Basic Usage

import urllib.request

import skimage
import torch
import torch.nn.functional as F
import torchvision
import torchvision.transforms

import torchxrayvision as xrv

model_name = "densenet121-res224-chex"

img_url = "https://huggingface.co/spaces/torchxrayvision/torchxrayvision-classifier/resolve/main/16747_3_1.jpg"
img_path = "xray.jpg"
urllib.request.urlretrieve(img_url, img_path)

model = xrv.models.get_model(model_name, from_hf_hub=True)

img = skimage.io.imread(img_path)
img = xrv.datasets.normalize(img, 255)

# Check that images are 2D arrays
if len(img.shape) > 2:
    img = img[:, :, 0]
if len(img.shape) < 2:
    print("error, dimension lower than 2 for image")

# Add color channel
img = img[None, :, :]

transform = torchvision.transforms.Compose([xrv.datasets.XRayCenterCrop()])

img = transform(img)

with torch.no_grad():
    img = torch.from_numpy(img).unsqueeze(0)
    preds = model(img).cpu()
    output = {
        k: float(v)
        for k, v in zip(xrv.datasets.default_pathologies, preds[0].detach().numpy())
    }
print(output)

For more code examples, we refer to the example scripts.

📄 License

This project is licensed under the Apache 2.0 license.

📚 Documentation

Citation

Primary TorchXRayVision paper: https://arxiv.org/abs/2111.00595

Joseph Paul Cohen, Joseph D. Viviano, Paul Bertin, Paul Morrison, Parsa Torabian, Matteo Guarrera, Matthew P Lungren, Akshay Chaudhari, Rupert Brooks, Mohammad Hashir, Hadrien Bertrand
TorchXRayVision: A library of chest X-ray datasets and models. 
https://github.com/mlmed/torchxrayvision, 2020


@article{Cohen2020xrv,
author = {Cohen, Joseph Paul and Viviano, Joseph D. and Bertin, Paul and Morrison, Paul and Torabian, Parsa and Guarrera, Matteo and Lungren, Matthew P and Chaudhari, Akshay and Brooks, Rupert and Hashir, Mohammad and Bertrand, Hadrien},
journal = {https://github.com/mlmed/torchxrayvision},
title = {{TorchXRayVision: A library of chest X-ray datasets and models}},
url = {https://github.com/mlmed/torchxrayvision},
year = {2020}
arxivId = {2111.00595},
}

and this paper which initiated development of the library: https://arxiv.org/abs/2002.02497

Joseph Paul Cohen and Mohammad Hashir and Rupert Brooks and Hadrien Bertrand
On the limits of cross-domain generalization in automated X-ray prediction. 
Medical Imaging with Deep Learning 2020 (Online: https://arxiv.org/abs/2002.02497)

@inproceedings{cohen2020limits,
  title={On the limits of cross-domain generalization in automated X-ray prediction},
  author={Cohen, Joseph Paul and Hashir, Mohammad and Brooks, Rupert and Bertrand, Hadrien},
  booktitle={Medical Imaging with Deep Learning},
  year={2020},
  url={https://arxiv.org/abs/2002.02497}
}