Update README.md

README.md

@@ -1,13 +1,54 @@
 ---
 license: bigcode-openrail-m
 language:
-- en
-- fr
+  - en
+  - fr
 metrics:
-- accuracy
+  - accuracy
 pipeline_tag: image-classification
 tags:
-- biology
-- code
-- tensorflow
----
+  - biology
+  - code
+  - tensorflow
+---
+
+# Mushroom Classification Model - JarviSpore
+
+This repository contains **JarviSpore**, a mushroom image classification model trained on a multi-class dataset with 23 different types of mushrooms. Developed from scratch with TensorFlow and Keras, this model aims to provide accurate mushroom identification using advanced deep learning techniques, including *Grad-CAM* for interpreting predictions. This project explores the performance of from-scratch models compared to transfer learning.
+
+## Model Details
+
+- **Architecture**: Custom CNN (Convolutional Neural Network)
+- **Number of Classes**: 23 mushroom classes
+- **Input Format**: RGB images resized to 224x224 pixels
+- **Framework**: TensorFlow & Keras
+- **Training**: Conducted on a machine with an i9 14900k processor, 192GB RAM, and an RTX 3090 GPU
+
+## Key Features
+
+1. **Multi-Class Classification**: The model can predict among 23 mushroom species.
+2. **Regularization**: Includes L2 regularization and Dropout to prevent overfitting.
+3. **Class Weighting**: Manages dataset imbalances by applying specific weights for each class.
+4. **Grad-CAM Visualization**: Utilizes Grad-CAM to generate heatmaps, allowing visualization of the regions influencing the model's predictions.
+
+## Model Training
+
+The model was trained using a structured dataset directory with data split as follows:
+- `train`: Balanced training dataset
+- `validation`: Validation set to monitor performance
+- `test`: Test set to evaluate final accuracy
+
+Main training hyperparameters include:
+- **Batch Size**: 32
+- **Epochs**: 20 with Early Stopping
+- **Learning Rate**: 0.0001
+
+Training was tracked and logged via MLflow, including accuracy and loss curves, as well as the best model weights saved automatically.
+
+## Model Usage
+
+### Prerequisites
+
+Ensure the following libraries are installed:
+```bash
+pip install tensorflow pillow matplotlib numpy