How anjaneyayashwanth/fruit-freshness-yolov8 works
This is a fruit-freshness classifier that labels images Fresh, Ripe, or Overripe and reports confidence, probabilities, and a freshness score. The main user workflow is image upload in the Streamlit app, image/texture preparation, selected-model inference, and displayed results. Separate scripts support dataset preparation, training and hyperparameter search, and model evaluation. The README documents both dual-stream and 5-channel models. Some listed modules were not excerpted, so their internal behavior and connections are left unspecified.
Main entry points: dataset_loader.py Split dataset loader · app.py Streamlit interface · dual_stream_model.py Dual-stream network · train_dual_stream.py Dual-stream training
Components 21
Prediction App
- Streamlit interface
app.py - Image preparation
app.py - Dual-stream inference
app.py - 5-channel inference
app.py - Prediction results
app.py - Single-image prediction
predict.py - Dual-stream prediction
predict_dual.py - Webcam prediction
webcam_predict.py
Data Preparation
- Dataset splitting
dataset_split.py - Dataset preprocessing
preprocessing.py - Split dataset loader
dataset_loader.py
Models and Training
- Dual-stream network
models/dual_stream_model.py - Dual-stream training
train_dual_stream.py - 5-channel training
train_custom.py - 5-channel tuning
gwo_train.py - Dual-stream tuning
gwo_dual_train.py
Evaluation
- Dual-stream evaluation
evaluate_dual.py - Dual-stream confusion matrix
confusion_matrix_dual.py - 5-channel confusion matrix
confusion_matrix.py
Other components
- Image uploader
- Ultralytics YOLO
Connections 19
- Image uploader → Streamlit interface: uploads image
app.py - Streamlit interface → Image preparation: prepares image
app.py - Streamlit interface → Dual-stream inference: selects model
app.py - Streamlit interface → 5-channel inference: selects model
app.py - Dual-stream inference → Dual-stream network: runs model
app.py - Streamlit interface → Prediction results: shows prediction
app.py - Dual-stream training → Split dataset loader: loads samples
train_dual_stream.py - Dual-stream training → Dual-stream network: trains model
train_dual_stream.py - 5-channel training → Split dataset loader: loads samples
train_custom.py - 5-channel training → Ultralytics YOLO: builds model
train_custom.py - 5-channel tuning → Split dataset loader: loads samples
gwo_train.py - Dual-stream tuning → Split dataset loader: loads samples
gwo_dual_train.py - Dual-stream tuning → Dual-stream network: evaluates parameters
gwo_dual_train.py - Dual-stream evaluation → Split dataset loader: loads test data
evaluate_dual.py - Dual-stream evaluation → Dual-stream network: evaluates model
evaluate_dual.py - Dual-stream confusion matrix → Split dataset loader: loads test data
confusion_matrix_dual.py - Dual-stream confusion matrix → Dual-stream network: gets predictions
confusion_matrix_dual.py - 5-channel confusion matrix → Split dataset loader: loads validation data
confusion_matrix.py - 5-channel confusion matrix → Ultralytics YOLO: builds model
confusion_matrix.py