- Created a tool to detect images of blood cells using tensorflow 2.3.0
- Engineered images for better fit.
- Used Pre-trained model (ReNet50) for predictions.
- Built a client facing API using Flask, HTML and Javascript.
- Python Version: 3.8.5
- Tensorflow Version : 2.3.0
- Packages: numpy, matplotlib, flask, tensorflow-2.0
- For Web Framework Requirements: pip install -r requirements.txt
- This Dataset is taken from the official NIH Website: https://www.kaggle.com/paultimothymooney/blood-cells
- Productionization: https://www.youtube.com/watch?v=BUh76-xD5qU
The diagnosis of blood-based diseases often involves identifying and characterizing patient blood samples. Automated methods to detect and classify blood cell subtypes have important medical applications.
- This dataset contains 12,500 augmented images of blood cells (JPEG) with accompanying cell type labels (CSV).
- There are approximately 3,000 images for each of 4 different cell types grouped into 4 different folders (according to cell type).
- The cell types are Eosinophil, Lymphocyte, Monocyte, and Neutrophil.
The model was build using Pre-Trained model ResNet50 https://keras.io/api/applications/resnet/#resnet50-function
In this step, I built a flask API endpoint with GUI that was hosted on a local webserver. The API endpoint takes in a request as image and returns a prediction.
