Updated "CBAM_ResNet50_Cervical_Classification" and made it deployment ready

.devcontainer/devcontainer.json

@@ -0,0 +1,33 @@
+{
+  "name": "Python 3",
+  // Or use a Dockerfile or Docker Compose file. More info: https://containers.dev/guide/dockerfile
+  "image": "mcr.microsoft.com/devcontainers/python:1-3.11-bookworm",
+  "customizations": {
+    "codespaces": {
+      "openFiles": [
+        "README.md",
+        "CBAM_ResNet50_Cervical_Classification/app.py"
+      ]
+    },
+    "vscode": {
+      "settings": {},
+      "extensions": [
+        "ms-python.python",
+        "ms-python.vscode-pylance"
+      ]
+    }
+  },
+  "updateContentCommand": "[ -f packages.txt ] && sudo apt update && sudo apt upgrade -y && sudo xargs apt install -y <packages.txt; [ -f requirements.txt ] && pip3 install --user -r requirements.txt; pip3 install --user streamlit; echo '✅ Packages installed and Requirements met'",
+  "postAttachCommand": {
+    "server": "streamlit run CBAM_ResNet50_Cervical_Classification/app.py --server.enableCORS false --server.enableXsrfProtection false"
+  },
+  "portsAttributes": {
+    "8501": {
+      "label": "Application",
+      "onAutoForward": "openPreview"
+    }
+  },
+  "forwardPorts": [
+    8501
+  ]
+}

.gitattributes

@@ -0,0 +1 @@
+CBAM_ResNet50_Cervical_Classification/cbam_resnet50_cervical/*.pth filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -190,6 +190,9 @@ dmypy.json
 *.pb
 *.tflite
 
+# Exception: Allow specific model file for deployment
+!CBAM_ResNet50_Cervical_Classification/cbam_resnet50_cervical/best_model.pth
+
 # IDE and Editor files
 .vscode/
 .idea/

.python-version

@@ -0,0 +1 @@
+3.11

CBAM_ResNet50_Cervical_Classification/.gitignore

@@ -0,0 +1,31 @@
+# Python
+__pycache__/
+*.py[cod]
+*$py.class
+*.so
+.Python
+env/
+venv/
+*.egg-info/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+*.ipynb
+
+# IDE
+.vscode/
+.idea/
+
+# OS
+.DS_Store
+Thumbs.db
+
+# Model outputs (keep best_model.pth)
+cbam_resnet50_cervical/*.csv
+cbam+resnet_outputs/*
+gradcam_outputs/*
+gradcam_specific_class_outputs/*
+
+# Temporary files
+*.log
+*.tmp

CBAM_ResNet50_Cervical_Classification/.python-version

@@ -0,0 +1 @@
+3.11

CBAM_ResNet50_Cervical_Classification/.streamlit/config.toml

@@ -0,0 +1,14 @@
+[theme]
+primaryColor = "#1f77b4"
+backgroundColor = "#FFFFFF"
+secondaryBackgroundColor = "#F0F2F6"
+textColor = "#262730"
+font = "sans serif"
+
+[server]
+maxUploadSize = 200
+enableCORS = false
+enableXsrfProtection = true
+
+[browser]
+gatherUsageStats = false

CBAM_ResNet50_Cervical_Classification/README.md

@@ -1,116 +1,106 @@
-# CBAM-ResNet50 Cervical Classification
+# CBAM-ResNet50 Cervical Cancer Cell Classifier
 
-This folder contains the code and notebooks for training, evaluating and explaining a CBAM-ResNet50 model for cervical cell image classification. The language used here is intentionally simple so that both technical and non-technical readers can follow the workflow.
+🔬 **Complete Pipeline**: Raw Image → Preprocessing → Classification → Explainability
 
-## Folder Contents
+## Features
 
-- `CBAM_ResNet50_Cervical_Classification.ipynb` – main training and evaluation notebook
-- `CBAM_ResNet50_Explainability_GradCAM.ipynb` – explainability and visualisation notebook
-- `app.py` – script for running the trained model in an application/dashboard setting
-- `cbam_resnet50_cervical/` – saved model weights and metric outputs
-- `gradcam_outputs/`, `gradcam_specific_class_outputs/` – saved Grad-CAM heatmaps
-- `sample_image/` – example image used for demos
-- `requirements.txt` – Python dependencies for this module
-- `README_DASHBOARD.md` – separate documentation for the dashboard application
+- **Advanced Model Architecture**: CBAM-ResNet50 with Channel & Spatial Attention
+- **Preprocessing Pipeline**: Resize → NLM Denoising → CLAHE Enhancement
+- **Explainable AI**: GradCAM++ visualization
+- **Interactive Dashboard**: Built with Streamlit
+- **High Performance**: 94.32% accuracy on test set
 
----
+## Cell Types Classified
 
-## 1. CBAM_ResNet50_Cervical_Classification.ipynb
+1. Dyskeratotic
+2. Koilocytotic
+3. Metaplastic
+4. Parabasal
+5. Superficial-Intermediate
 
-This is the main notebook where the CBAM-ResNet50 model is built, trained and evaluated on cervical cell images.
+## Quick Start
 
-At a high level, the notebook walks through:
+### Local Installation
 
-1. **Project overview and motivation**  
-   Explains why cervical cell classification is important and how attention mechanisms (CBAM) can help the model focus on the nucleus and other relevant regions.
-
-2. **Dataset setup**  
-   - Describes the folder structure of the dataset (one folder per cell type).  
-   - Shows how image paths and labels are loaded.  
-   - Applies standard image preprocessing and augmentation (resize, normalization, flips, etc.).
-
-3. **CBAM-ResNet50 model definition**  
-   - Starts from a ResNet50 model pretrained on ImageNet.  
-   - Adds Convolutional Block Attention Modules (CBAM) that learn **what** features and **where** in the image to focus on.  
-   - Modifies the final layer to predict the five cervical cell classes.
-
-4. **Training pipeline**  
-   - Defines loss function (for example cross-entropy) and optimizer.  
-   - Sets training hyperparameters such as learning rate, batch size and number of epochs.  
-   - Trains the model over multiple epochs, printing progress for training and validation.
-
-5. **Evaluation and metrics**  
-   - Computes overall accuracy and per-class accuracy.  
-   - Saves detailed per-class metrics to `per_class_metrics.csv`.  
-   - Plots and saves a confusion matrix (`confusion_matrix.png`) and training curves (`training_history.png`).
-
-6. **Saving results**  
-   - Saves the best-performing model weights to `cbam_resnet50_cervical/best_model.pth`.  
-   - Exports predictions to `test_predictions.csv`.  
-   - Stores training history in `training_history.csv` for later analysis.
-
-If you want to reproduce the training, open this notebook, adjust the dataset paths to your environment, install the dependencies listed in `requirements.txt`, and run the cells from top to bottom.
-
----
-
-## 2. CBAM_ResNet50_Explainability_GradCAM.ipynb
-
-This notebook focuses on making the CBAM-ResNet50 model more interpretable using Grad-CAM based visual explanations.
+```bash
+# Clone the repository
+git clone <your-repo-url>
+cd CBAM_ResNet50_Cervical_Classification
 
-The main steps are:
+# Install dependencies
+pip install -r requirements.txt
 
-1. **Environment setup**  
-   - Installs and imports the required libraries, including `pytorch-grad-cam`, `torch`, `torchvision`, `opencv-python` and `matplotlib`.  
-   - Detects whether the notebook is running in Google Colab or locally.
+# Run the app
+streamlit run app.py
+```
 
-2. **Model reconstruction and weight loading**  
-   - Re-creates the same CBAM-ResNet50 architecture used during training.  
-   - Loads the trained weights from `cbam_resnet50_cervical/best_model.pth`.  
-   - Moves the model to CPU or GPU depending on availability.
+### Streamlit Cloud Deployment
 
-3. **Image preprocessing**  
-   - Defines transforms to resize, normalize and convert images to tensors in the same way as during training.  
-   - Loads either a `sample_image` from this folder or user-specified test images.
+1. Push your code to GitHub
+2. Go to [share.streamlit.io](https://share.streamlit.io)
+3. Click "New app"
+4. Select your repository
+5. Set main file path: `CBAM_ResNet50_Cervical_Classification/app.py`
+6. Click "Deploy"
 
-4. **Grad-CAM and Grad-CAM++ generation**  
-   - Sets up Grad-CAM / Grad-CAM++ on the chosen model layers.  
-   - Generates class-specific heatmaps that highlight which parts of the image influenced the model’s decision.  
-   - Overlays these heatmaps on the original images for intuitive visualisation.
+**Important**: Ensure these files are in your repository:
+- `cbam_resnet50_cervical/best_model.pth` (trained model)
+- `sample_image/original_images/` (sample images)
+- `requirements.txt`
+- `packages.txt`
+- `.streamlit/config.toml`
 
-5. **Saving and organising outputs**  
-   - Saves attention and Grad-CAM maps into `gradcam_outputs/`.  
-   - Optionally organises examples per class under `gradcam_specific_class_outputs/`.  
-   - Helps you visually inspect how the model attends to different cervical cell types.
+## Usage
 
-You can use this notebook to understand and validate the behaviour of the trained model, especially whether it is focusing on the nucleus and other medically relevant regions.
+### Upload Image
+1. Use the sidebar to upload a cervical cell microscopy image
+2. Supported formats: BMP, PNG, JPG, JPEG
 
----
+### Use Sample Images
+1. Select "Use Sample Images" in the sidebar
+2. Choose from pre-loaded sample images
+3. View instant results
 
-## 3. app.py
+### View Results
+- **Step 1**: Image preprocessing visualization
+- **Step 2**: Classification results with confidence scores
+- **Step 3**: GradCAM++ explainability heatmaps
 
-`app.py` is a Python script intended to expose the trained CBAM-ResNet50 model as an interactive application (for example, a web dashboard). While details are documented in `README_DASHBOARD.md`, at a high level it typically:
+### Download Results
+- Preprocessed image
+- GradCAM++ visualization
+- Complete results (JSON)
 
-- Loads the trained model weights from `cbam_resnet50_cervical/best_model.pth`.  
-- Preprocesses uploaded cervical cell images to match the training pipeline.  
-- Runs inference and returns the predicted cell class and confidence scores.  
-- Optionally displays attention maps or Grad-CAM overlays to improve interpretability for end users.
+## Model Performance
 
-This script is useful if you want to present the model to clinicians or other stakeholders in a user-friendly way without requiring them to run the notebooks.
+| Metric | Score |
+|--------|-------|
+| Accuracy | 94.32% |
+| Precision | 94.40% |
+| Recall | 94.32% |
+| F1-Score | 94.34% |
 
----
+## Technical Details
 
-## 4. How to Get Started
+- **Parameters**: 24,214,989
+- **Training Dataset**: 4,049 images (3,239 train / 405 val / 405 test)
+- **Preprocessing**: NLM (h=3), CLAHE (clipLimit=1.2)
+- **Input Size**: 224×224
+- **Framework**: PyTorch
 
-1. **Install dependencies** using the provided `requirements.txt` (ideally inside a virtual environment):
+## Dependencies
 
-```bash
-pip install -r requirements.txt
-```
+- Python >= 3.8
+- PyTorch >= 2.0.0
+- Streamlit >= 1.28.0
+- OpenCV (headless for deployment)
+- scikit-image
+- plotly
 
-2. **Run the training notebook** `CBAM_ResNet50_Cervical_Classification.ipynb` to train or fine-tune the model on your dataset, or load the already provided `best_model.pth` if you only want to run inference.
+## Notes
 
-3. **Use the explainability notebook** `CBAM_ResNet50_Explainability_GradCAM.ipynb` to generate visual explanations for model predictions.
+⚠️ **Disclaimer**: This tool is for research purposes only. Always consult medical professionals for diagnosis.
 
-4. **Launch the application** using `app.py` (see `README_DASHBOARD.md` for specific run instructions) to provide an interactive interface for predictions.
+## License
 
-This structure allows you to go from raw cervical cell images to a trained attention-based classifier, interpret its decisions, and deploy it in an application-oriented setting from within this single folder.
+This project is for educational and research purposes.