InuriDinethma/black-tea-powder-classifier

Black Tea Powder Classifier

This repository contains a Gradio app and the trained EfficientNetV2S Keras model for black tea powder classification. It also supports open-set rejection through the NOT_TEA class.

The app loads these files from the Space:

black_tea_efficientnetv2s_crop068_best.keras
class_names.json
preprocessing_config.json

The model was trained with an image size of 384, a central crop fraction of 0.68, and ten-crop test-time augmentation. The model predicts these nine classes:

BM, BOP, BP, BROKEN_TEA, DUST, FANNING_2, PF, PW_DUST, NOT_TEA

Run Locally

Install the dependencies:

pip install -r requirements.txt

Start the Gradio app:

python app.py