NITANSH THAKERNT / SYSTEMS
Developer
Independent project

Automated Wordle Solver with Visual Perception

A real-time assistant combining board detection, OCR, colour classification, and heuristic action selection.

All work
Visual-perception pipeline from screenshot to extracted state and recommendationA Wordle-recognition pipeline. A cobalt-blue line scans the grid for OCR. Green tiles mean a letter is correct and in the correct position; amber tiles mean a letter is present in the wrong position; neutral tiles are unknown or absent. The interpreted colour state informs a heuristic action.SCAN / OCRCRANESHEARPRONEOCR + COLOURSTATEHEURISTICACTION
Image / extracted state / recommendationArchitecture abstraction

The problem

Recover reliable game state from screenshots before recommending the next action.

System architecture

  • OpenCV board detection and OCR using Tesseract and EasyOCR.
  • Colour classification and state extraction feeding heuristic word filtering.
  • A Streamlit interface for live feedback.

Engineering decisions

  • Separate perception from recommendation logic.
  • Visualise extracted state so recognition errors remain visible.

Testing and verification

  • Screenshot-based board and state checks.

Limitations

  • OCR and colour classification depend on capture quality and theme.

What I would improve next

  • Broaden visual-theme robustness and confidence reporting.