README.md

metadata

title: Bird vs Drone Classification
emoji: 🦅🛸
colorFrom: blue
colorTo: indigo
sdk: docker
app_port: 7860
pinned: false

Bird vs Drone Image Classification

An end-to-end deep learning project to classify airborne objects into "Bird" or "Drone" categories using a Convolutional Neural Network (MobileNetV2).

Features

• Deep Learning Model: MobileNetV2 based architecture for fast and accurate classification.
• Data Pipeline: Automated conversion from YOLO detection labels to classification datasets.
• Web Interface: Premium glassmorphic UI for real-time inference.
• Data Augmentation: Robust training using rotation, flip, and zoom augmentations.

Project Structure

• prepare_data.py: Prepares the dataset manifests.
• train_model.py: Trains the model on a subset of the 20k+ images.
• app.py: Flask application for the web interface.
• templates/ & static/: Frontend assets.
• bird_vs_drone_model.h5: The trained model weights.

Installation

pip install -r requirements.txt

Usage

1. Prepare Data:

   python prepare_data.py
   

2. Train Model:

   python train_model.py
   

3. Run Web App:

   python app.py
   

Results

The system provides a confidence score and a visual analysis of the uploaded target, distinguishing between natural avian flight and synthetic drone movement.