Amazon Explorer

The Blogger that tells you everything you need to know about Amazon Rainforest

Comment

Inspiration

The Amazon rainforest holds secrets of ancient civilizations, hidden beneath its dense canopy for centuries. Inspired by recent discoveries of pre-Columbian geoglyphs, engineered soil (terra preta), and advanced settlement planning, I aimed to build a tool that combines satellite imagery and AI to help archaeologists, researchers, and curious explorers identify areas of potential historical significance. I was driven by the question: What if the legends of cities like El Dorado weren’t myths—but forgotten truths waiting to be uncovered from space?

What it does

Amazon Explorer is an AI-powered visual analysis platform that scans satellite imagery to highlight regions with patterns or features that may indicate human activity—such as circular clearings, road systems, soil anomalies, and other geo-structural clues. It also let's you to discover places in the Amazon.

It generates:

AI-driven interpretations of satellite images. Hypotheses about what these patterns may represent (e.g., geoglyphs, ancient agricultural grids, etc.). How we built it Satellite Imagery: I sourced images from the Google Map Platform's Static API. AI Interpretation: I used Google Gemini to analyze satellite data, detect unusual patterns, and generate historical hypotheses. Interface: Built a simple web-based dashboard for viewing AI-generated insights.

Challenges we ran into:

  1. Cloud Obstruction: Many Amazon images had heavy cloud cover, making analysis difficult.
  2. False Positives: Distinguishing ancient earthworks from modern deforestation or logging roads required refined logic and multiple validation layers.
  3. Data Volume: Processing large multi-band satellite images was computationally intensive.
  4. Historical Ambiguity: Training the AI to understand what a “lost civilization site” might look like from space required carefully curated prompts and examples.

Accomplishments that we're proud of

  1. We successfully identified multiple regions that visually align with known geoglyphs and pre-Columbian settlement zones.
  2. We created a working AI model that can describe satellite imagery in archaeological terms.
  3. We built a tool that could potentially assist real-world researchers in narrowing down areas for field study.

What we learned

  1. How to effectively reveal hidden features in dense forests.
  2. How Gemini can be guided to interpret remote sensing data meaningfully.
  3. That the combination of AI + Earth observation is a powerful approach to uncovering human history.

What's next for Amazon Explorer

  1. Automated Scanning: Build a pipeline that continuously scans new tiles of the Map Platform's Map tiles API over the Amazon and flags anomalies.
  2. Expansion: Apply this model to other underexplored areas like the Congo Basin or Southeast Asia.

Built With

Share this project:

Updates