https://sheldonsebastian.github.io/Red-Blood-Cell-Classification/
| Path | Description |
|---|---|
| /POCs | Files containing the proof of concepts/experimentation |
| /docs | Files related to report website |
| /input | Training image data and labels split into train, validation and holdout set |
| /src/common | Common utility functions used by all scripts |
| /src/model_trainers | Files containing code for training the model using: 1. manual hyperparameter tuning 2. random search hyperparameter tuning 3. Optuna (Automatic hyperparameter tuning) |
| /src/0_preprocess.py | Code to preprocess the image files and split into train-validation-holdout splits |
| /src/1_eda.py | Exploratory Data Analysis Jupyter Notebook |
| /src/2_inference.py | Using the trained models make inference on validation and holdout set |
| requirements.txt | List of all the packages used for this project |
- Download repository
- To install all required python packages use: conda create --name rbc_classification --file requirements.txt
- Update BASE_DIR in src/0_manual.py, src/1_random_search.py, src/2_hyper_optimizer.py to current directory on your machine
- Run src/0_manual.py, src/1_random_search.py, src/2_hyper_optimizer.py to train models and save them in saved_models directory
- Run src/2_inference_holdout.py to perform inference on holdout(unseen) data.