This project is part of a task for the college where I study, so
task-parts
contains files that associated with that task, whishing that I would get the full mark ;). In general the base code doesn't have any special parts except that folder.
After cloning the repository, install the required packages in a virtual environment.
Next, download the datasets and checkpoints, as describe below.
- Download the Chen et al. labels and the chest X-rays in png format for IU X-Ray from:
https://openi.nlm.nih.gov
- Place the files into
dataset
folder, such that their paths aredataset/reports
anddataset/images
.
This approach uses CheXNet
, and DenseNet121
as a CNN Encoder model. By default the CheXNet
pretrained weights are located in weights
folder.
The model configurations for each task can be found in its config.py
file.
Use the below command to train the model form a saved checkpoint or without a checkpoint.
python train.py
The model performance measure is based of the BLEU
metric.
Feel free to change the performance measure metric in the
check_accuracy
method that is located in theeval.py
file
Run the following command to calculate BLEU
score.
python eval.py