Update README.md

README.md

@@ -1,12 +1,128 @@
+# 🏛️ Landmark Detection using VGG-19 (Transfer Learning)
+
+This project implements an end-to-end **Landmark Image Classification system** using a pre-trained **VGG-19 Convolutional Neural Network** and deploys it as a web application using **Gradio** on **Hugging Face Spaces**.
+
+The model is trained on a subset of the **Google Landmarks Dataset v2 (Top 51 Categories)** and is capable of predicting the landmark class of an input image.
+
+---
+
+## 🚀 Features
+
+* Automatic dataset loading from Hugging Face (`pemujo/GLDv2_Top_51_Categories`)
+* Image preprocessing (resizing, normalization)
+* Transfer Learning using **VGG-19 pretrained on ImageNet**
+* Custom classification head for landmark recognition
+* Training and evaluation on CPU
+* Interactive **Gradio Web Interface** for real-time prediction
+
 ---
-title: Landmark Detection
-emoji: 📚
-colorFrom: blue
-colorTo: red
-sdk: gradio
-sdk_version: 6.3.0
-app_file: app.py
-pinned: false
+
+## 🧠 Model Architecture
+
+* Base Model: VGG-19 (without top classification layers)
+* Frozen convolutional layers for feature extraction
+* Custom layers:
+
+  * Flatten
+  * Dense (ReLU)
+  * Dropout
+  * Softmax output layer (51 classes)
+
+---
+
+## 📂 Dataset
+
+* Source: Google Landmarks v2 (Top 51 Categories)
+* Platform: Hugging Face Datasets
+* Classes: 51 landmark categories
+* Format: Image + Numeric Label
+
+Due to numeric class encoding, outputs are displayed as:
+
+```
+Landmark_0, Landmark_1, ..., Landmark_50
+```
+
+---
+
+## 🛠️ Tech Stack
+
+* Python
+* TensorFlow / Keras
+* Hugging Face Datasets
+* Gradio
+* NumPy
+* Pillow
+
+---
+
+## ▶️ How It Works
+
+1. Dataset is downloaded automatically using Hugging Face `datasets` library.
+2. Images are resized to 224×224 and normalized.
+3. VGG-19 extracts deep visual features.
+4. Custom classifier predicts the landmark category.
+5. Gradio provides a web interface for uploading and classifying images.
+
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+## 📦 Installation (Handled Automatically in Space)
+
+Dependencies are listed in `requirements.txt`:
+
+```
+tensorflow
+datasets
+pillow
+matplotlib
+gradio
+```
+
+---
+
+## 🖼️ Web App Usage
+
+1. Upload a landmark image (JPEG/PNG).
+2. The model predicts the most probable landmark class.
+3. Output is shown as `Landmark_X`.
+
+---
+
+## 📈 Sample Output
+
+```
+Prediction: Landmark_43
+```
+
+This indicates the input image belongs to the 44th landmark category learned by the model.
+
+---
+
+## 📚 Academic Use
+
+This project demonstrates:
+
+* Transfer Learning
+* Deep CNN Feature Extraction
+* Image Classification Pipeline
+* Model Deployment with Gradio
+* Reproducible ML using Hugging Face Hub
+
+---
+
+## 👨‍🎓 Author
+
+**Apurba Das**
+Landmark Detection Project
+Deep Learning & Computer Vision
+VGG-19 | TensorFlow | Hugging Face | Gradio
+
+---
+
+## 🔮 Future Improvements
+
+* Map numeric labels to actual landmark names
+* Fine-tune deeper VGG-19 layers
+* Add top-k prediction probabilities
+* Deploy with GPU for faster training
+* Expand to full Google Landmarks v2 dataset