🌲 Forest Watch

🌲 About Forest Watch

AI-powered deforestation detection for Sumatra — analyze satellite or aerial images to identify active forest clearing and support conservation efforts.

Project overview

Forest Watch analyzes aerial and satellite images (Google Earth Pro) to automatically detect active deforestation — areas showing clearing, exposed soil, or logging patterns. The system focuses on timely, actionable detections to help researchers and policy makers.

Dataset collection

Imagery captured at various altitudes and locations across Sumatra. Samples were filtered for clouds, blur, and duplicates before annotation.

  • Source: Google Earth Pro
  • Top-down, north-oriented images
  • Camera altitude: 1000m – 2000m (close-ups: 100–300m)

Dataset statistics

Collected
132
After filtering
122
Annotated
61

Filtering removed images with heavy cloud cover, blur, inconsistent lighting, or duplicates.

Detection criteria

🟤

Color indicators

Brown, gray, or yellowish areas lacking green vegetation.

📐

Geometric patterns

Open clearings with rectilinear or repetitive clearing shapes.

🪵

Logging evidence

Visible logging roads, stacked logs, or equipment traces.

🌲

Forest context

Located inside or adjacent to forest cover (not plantations or paddy fields).

Note: Palm oil plantations, mines, rice paddies, and residential areas are excluded from deforestation classification.

Technology stack

Data collection

  • • Google Earth Pro
  • • Manual annotation & filtering

AI & training

  • • Roboflow (training & deployment)
  • • Object detection workflow

Project goal

Provide an accessible automated tool to monitor deforestation in Sumatra, enabling timely intervention and conservation.