README.md

metadata

title: Analogic Watch Detector
emoji: ⌚
colorFrom: red
colorTo: gray
sdk: gradio
sdk_version: 5.9.1
app_file: app.py
pinned: false
license: mit
short_description: Automatic time reading from analog clock images using AI.

Analog Watch Detector

A computer vision application that automatically reads time from images of analog clocks. This project uses a YOLOv8 model to detect clock components (hands, center, digits) and geometrically calculates the time.

🚀 Application Features

• Time Detection: Upload an image of a watch or clock, and the AI will predict the time.
• Visual Feedback: Displays the detected components (Hour hand, Minute hand, Second hand) overlaid on the image.
• Test Examples: Includes a set of test images to try out immediately.

🛠️ How it Works

The model detects:

1. Clock Face/Circle
2. Hour Hand
3. Minute Hand
4. Second Hand (if present)
5. The Number 12 (for orientation)

Using the coordinates of these components, the application calculates the angles of the hands relative to the center and the 12 o'clock position to determine the exact time.

📦 Dataset & Model

• Model: YOLOv8s custom trained on a diverse dataset of watches and clocks.
• Training: Optimized using hyperparameter tuning (tune4 - best.pt).

👨‍💻 Credits

Developed by Ana Pinto and Pedro Leitão. Project developed for a Computer Vision course.