/ml_chest_xray_issue_classifaction

Anomaly detection on human chest X-ray images

Primary LanguageJupyter NotebookMIT LicenseMIT

Chest abnormalities detection

License Python 3.7 scikit-learn 0.23.2 Solvve

Description

Detection of abnormalities on chest X-ray images.

We need to find several abnormalities on chest X-ray images.

Example of images

example

For solving this problem we will apply FasterRCNN, Yolo_v5 using PyTorch and PyTorch_Lightning libraries

And additionally we will build binary classifier to classify image - normal/abnormal

As a result of our work we will build simple web app using Flask which is capable of taking image and detecting abnormalities on it.

Installation yolo v5

For install Yolo v5 run following commands:

  1. !git clone https://github.com/ultralytics/yolov5
  2. !mv yolov5/* ./
  3. !pip install -r requirements.txt

Analysis and modeling

We follow the next steps:

  1. EDA (notebooks/EDA.ipynb)
  2. Modeling : Yolo v5 (notebooks/Yolov5.ipynb)
  3. Modeling : FasterRCNN (notebooks/FasterRCNN.ipynb)
  4. Modeling : binary classifier - abnormal/normal image (notebooks/binary_classifier.ipynb)

Datasets

  1. https://www.kaggle.com/awsaf49/vinbigdata-512-image-dataset

  2. https://www.kaggle.com/raddar/vinbigdata-competition-jpg-data-2x-downsampled

  3. https://www.kaggle.com/c/vinbigdata-chest-xray-abnormalities-detection/data

Web app for abnormaly detection

Example of web app for abnormalities detection

Step 1. Uploading image

upload

Step 2. Click submit and see the result

upload

upload

upload