This repository contains a multi-task deep learning model for COVID-19 classification, lung cancer detection, and lung segmentation. The model is based on the ResNet-50 backbone and utilizes convolutional neural networks (CNNs) for the individual tasks. In a multi-task learning setting, the significance of jointly training these tasks is to improve generalizability, learn efficient feature extraction, and effective usage of lesser data.
git clone https://github.com/AmiteshBadkul/LCFNN.git
cd LCFNN/environment/
conda env create -f environment.yaml
conda activate multi-task-learning
classification_dataset
directory. Here is the link to the dataset --> COVID19cancer_detection_dataset
directory. Here is the link to the dataset --> Lung Cancer Detectionsegmentation_dataset
directory. Here is the link to the dataset --> Lung Segmentationpython main.py
The project has the following structure:
code/
- Contains the code for the model for training and evaluation.results/
- Contains the results of the trained models as well as the model.analysis/
- Contains jupyter notebooks for analysis of the results obtained.The results currently obtained are baseline results more model improvements will improve the performance further.
The model achieves the following performance on the test set:
COVID-19 Classification:
Lung Cancer Detection:
Lung Segmentation:
Some notes and to-do list: