buiquangchien_6_readme.txt

1. TEAM MEMBERS & TASKS (FINAL)
- Bui Quang Chien (Team lead - 23001837): Implement extended classical ML pipeline, set up Deep Learning models (MobileNetV2, ResNet50, EfficientNetB0), and prepare the LaTeX report.
- Nguyen Anh Duc (23001865): Prepare LaTeX slides, presentation report, and develop a GUI application to integrate classification models (PyQt5/Tkinter).
- Nguyen Le Ngoc Bao (23001832): Model evaluation, hyperparameter tuning, error analysis (confusion matrix), and compare performance between classical ML and Deep Learning.

2. DATA DOWNLOAD
- Dataset used: TrashNet.
- Download link (zip): https://drive.google.com/file/d/15xpVIqcPHj6A_09JOSF3qz5E0cmonHjt/view?usp=sharing

3. FOLDER LAYOUT & EXPERIMENTAL WORKFLOW
- Example extracted dataset layout:

```
/dataset/
  /dataset-resized/
    /cardboard
    /glass
    /metal
    /paper
    /plastic
```

Experiment steps (final project):

- Step 1: Upload `buiquangchien_6_waste_classification.ipynb` to Google Colab and set runtime to a GPU (Runtime > Change runtime type > T4 GPU).
- Step 2: Run the first cells to mount Google Drive and extract the dataset. Ensure the dataset structure is under `/content/dataset/dataset-resized`.
- Step 3: Execute Phases 1–9 to evaluate classical ML pipelines.
- Step 4: Execute PART 2 and PART 3 to test new feature extraction and train hybrid/Deep Learning models (MobileNetV2, ResNet50, EfficientNetB0).
- Step 5: Model `.keras` files and evaluation plots will be saved to Google Drive under `Lastterm_Waste_Classification`. Use these `.keras` files for local GUI integration.