DeepChest - Efficient Deep Learning Framework for Detection of Chest Pathologies using Chest X-ray Images
The project uses the NIH Chest X-ray Dataset here is the here to acces it. This dataset has 112k images of Chest Xrays and these include the following diseases.
- Atelectasis
- Consolidation
- Infiltration
- Pneumothorax
- Edema
- Emphysema
- Fibrosis
- Effusion
- Pneumonia
- Pleural_thickening
- Cardiomegaly
- Nodule Mass
- Hernia
Through this project, we aim to enable low-power portable healthcare diagnostic support solutions. We explore Binarized Neural Networks (BNN) for the efficient diagnosis of thoracic pathologies via Chest X-Ray images. We test our model on the publicly available NIH Chest X-Ray dataset and achieve state-of-the-art results while consuming substantially less resources than the current state-of-the-art network, CheXNet.
Binary and Full Precision weights can be found here
Label: Cardiomegaly
Model Output: Cardiomegaly
Label: Cardiomegaly and Emphysema
Model Output: Cardiomegaly and Emphysema
Label: No Finding
Model Output: No Finding
- Place all the images in data folder
- Place train.csv and dev.csv in the same folder
- Run the following command
python test.py