Monument Recognition Model
Detailed information about our deep learning model for Indian monument recognition, including architecture, training data, and performance metrics.
Monument Recognition Model Information
Model Architecture
Base Model:
ResNet50
Pre-training:
ImageNet
Input Size:
224×224×3
Output Classes:
37 monuments
Model File:
monument_classifier_v2.h5
Web Format:
model.json + *.bin
Training Dataset
Total Images:
3,700+ images
Images per Monument:
100+ images
Image Resolution:
800×800 minimum
Data Split:
80% train, 20% val
Augmentation:
Rotation, Zoom, Brightness
Expected Performance
92%
Validation Accuracy
98%
Top-3 Accuracy
2-3s
Inference Time
15MB
Model Size
Technical Implementation
Training Configuration:
- • Optimizer: Adam with learning rate scheduling
- • Batch Size: 32
- • Epochs: 50 + 20 fine-tuning
- • Loss: Categorical Crossentropy
- • Early Stopping: Patience 10
Deployment:
- • Framework: TensorFlow.js
- • Runtime: Browser-based inference
- • Format: JSON + Binary weights
- • Preprocessing: Resize, normalize
- • Postprocessing: Softmax confidence
Monument Categories (37 Classes)
UNESCO World Heritage (10)
- • Taj Mahal
- • Red Fort
- • Qutub Minar
- • Ajanta Caves
- • Ellora Caves
- • Khajuraho Temples
- • Konark Sun Temple
- • Sanchi Stupa
- • Hampi Group
- • Pattadakal Group
Karnataka Heritage (13)
- • Mysore Palace
- • Virupaksha Temple
- • Stone Chariot Hampi
- • Belur Chennakeshava
- • Halebidu Hoysaleswara
- • Badami Caves
- • Aihole Temples
- • Gol Gumbaz
- • Shravanabelagola
- • Chitradurga Fort
- • Bangalore Fort
- • Tipu Palace
- • Melukote Temple
Other Major Sites (14)
- • Golden Temple
- • Gateway of India
- • Hawa Mahal
- • Charminar
- • Victoria Memorial
- • India Gate
- • Meenakshi Temple
- • Brihadeeswara Temple
- • Lotus Temple
- • Akshardham
- • Fatehpur Sikri
- • Amber Fort
- • Humayun's Tomb
- • Murudeshwar Temple
Model Files for Presentation:
Training Model:
monument_classifier_v2.h5
Web Model Architecture:
model.json
Web Model Weights:
*.bin files
Class Labels:
classes.json