YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
🏛 Architecture Style Bot
A Telegram bot that identifies the architectural style of a building from a photo.
@arc_style_bot — try the bot
How it works
Send the bot a photo of a building — it will identify the architectural style and return the top 3 predictions with confidence scores.
Before classification, the bot checks whether the photo actually contains a building using a MobileNetV2-based detector. If no building is found, it asks for a better photo.
Supported styles (15 classes)
Ancient Egyptian · Art Deco · Baroque · Brutalist · Byzantine · Classic · Colonial · Deconstructivism · Gothic · International · Modern · Postmodern · Romanesque · Russian Revival · Stalinist
Model
- Architecture: ResNet18, fine-tuned on custom data
- Dataset: International Architectural Styles Combined (Kaggle)
- Training: 17 epochs, backbone unfrozen at epoch 10
- Validation accuracy: ~65%
Features
- Preprocessing pipeline optimized for phone photos: EXIF orientation fix, edge cropping, sharpness and contrast enhancement
- Building detector filters out irrelevant images before classification
Stack
- Python 3.x
- PyTorch + torchvision
- pyTelegramBotAPI
- Pillow
Running locally
pip install -r requirements.txt
Create a .env file:
BOT_TOKEN=your_token_here
python main.py
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

