🌤️ Sky Segmentation & Condition Classification

A web-based AI application designed to detect and segment the sky region from any image. Powered by the Ultralytics YOLO API for precise instance segmentation, this project also features a custom JavaScript image processing algorithm to classify the sky condition into Clear (Cerah) or Mendung (Overcast) based on color analytics ($Blue Dominance$ and $Color Gap$).

🚀 Features

• AI-Powered Segmentation: Uses a state-of-the-art YOLO segmentation model to map out the exact boundaries of the sky.
• Intelligent Color Analysis: Built-in algorithm that bypasses camera auto-white balance traps by measuring pixel-level RGB variance.
• Binarized Classification: Automatically determines whether the sky is Cerah (Clear Blue) or Mendung (Cloudy/Overcast).
• Real-time Canvas Rendering: Draws high-fidelity colored overlay masks and bounding boxes directly over the detected areas.
• Developer Debug Mode: Console logging features that print real-time RGB averages, Brightness, and Color Gap for algorithm fine-tuning.

🛠️ Tech Stack

• Frontend: HTML5, CSS3, Bootstrap 5 (Custom Dark Theme Layout)
• Core Logic: Vanilla JavaScript (Canvas API & ImageData processing)
• AI Model: Ultralytics YOLO Segmentation API via Fetch requests

📸 Screenshots

🔧 How It Works (The Logic)

The script crops the bounding box provided by YOLO, samples the pixels, and runs the following mathematical evaluation:

1. Blue Dominance: $AvgB - \frac{AvgR + AvgG}{2}$ (Detects pure blue hues).
2. Color Gap: $Max(R,G,B) - Min(R,G,B)$ (Differentiates flat grey overcast from clear vibrant skies).

If BlueDominance > 50 and ColorGap > 50, the sky is flagged as Cerah, otherwise it defaults to Mendung to filter out camera sensor rona traps.

---

Developed with ❤️ by [Zudi Jago/ fzuhdi79-coder]