PATRYK KULPOK
← BACK TO SYSTEM MAPSYSTEM MAP ENTRY

SYSTEM S1

PUZZLEMIND

COMPUTER VISION / BOARD DETECTION / AUTOMATION R&D

Computer vision prototype for detecting and structuring a puzzle-game board from an emulator screen.

PYTHONOPENCVCOMPUTER VISIONIMAGE PROCESSINGAUTOMATIONWINDOWSBLUESTACKS

SYSTEM IDENTITY

SYSTEM IDPUZZLEMIND-S1

STATUSACTIVE R&D

TYPECOMPUTER VISION / AUTOMATION

Vision pipeline for locating a game board, validating its geometry and converting screen pixels into structured board data.

01 / OVERVIEW

PROJECT OVERVIEW

Program sees only pixels from the emulator window. To work on game state, the system must localize the correct board, transform it to a structured representation and validate candidate regions before downstream logic can act.

Core flow is intentionally narrow:

  • D-01

    EMULATOR SCREEN -> BOARD DETECTION

  • D-02

    BOARD REGION -> GRID

  • D-03

    GRID -> STRUCTURED STATE

PuzzleMind flow

  1. EMULATOR SCREEN
  2. VISION PIPELINE
  3. BOARD REGION
  4. GRID
  5. STRUCTURED STATE

02 / USE CASES

BOARD UNDERSTANDING USE CASES

UC-01

LOCATE THE BOARD

Find the puzzle board inside BlueStacks emulator content rather than relying on a fixed screen position.

UC-02

REJECT FALSE CANDIDATES

Filter regions that do not match expected board proportions or content boundaries.

UC-03

VALIDATE GEOMETRY

Check rectangularity, relative area and grid spacing before accepting a candidate.

UC-04

MAP AN 8X8 GRID

Convert the accepted board region into positional cells for downstream logic.

UC-05

EXPOSE BOARD STATE

Return structured cell and board metadata instead of raw pixels.

03 / VISUAL INPUT

VISUAL INPUT

Input is taken from BlueStacks content and should stay constrained to emulator content area. UI chrome and unrelated screen regions are filtered out before board extraction.

Input chain

  1. BLUESTACKS WINDOW
  2. CONTENT AREA
  3. SCREEN FRAME
  4. VISION PROCESSING

Window and search map

04 / BOARD DETECTION

BOARD DETECTION

A classical computer vision pipeline isolates candidate rectangles and prepares them for geometry checks.

Candidate generation

  1. SCREEN FRAME
  2. IMAGE PROCESSING
  1. CONTOUR CANDIDATES
  2. BOARD CANDIDATES

Candidate map with rejection markers

05 / GEOMETRY VALIDATION

GEOMETRY VALIDATION

Detection is only valid when candidate geometry satisfies multiple checks. This section acts as the quality gate before grid extraction.

CANDIDATE AACCEPT

Aspect ratio, rectangularity, content area and grid spacing passed.

  • ASPECT RATIO
  • RECTANGULARITY
  • RELATIVE AREA
  • CONTENT AREA
  • GRID SPACING

CANDIDATE BREJECT

Outside expected geometry for game board proportions.

CANDIDATE CREJECT

Irregular grid spacing compared with expected tile rhythm.

CANDIDATE REJECTION REASONS

  • implausible aspect ratio
  • insufficient rectangularity
  • outside BlueStacks content area
  • relative area outside configured range
  • irregular grid spacing

06 / GRID EXTRACTION

GRID EXTRACTION

Accepted board regions are projected to an 8x8 logical structure. The result is positional grid metadata that can be used by later logic.

Extraction flow

  1. BOARD REGION
  2. GRID STRUCTURE
  3. 8x8 CELL MAP
  4. STRUCTURED BOARD STATE

8x8 CELL MAP

0102030405060708
0102030405060708

CELL COORDINATES

07 / STRUCTURED OUTPUT

STRUCTURED OUTPUT

The objective is a structured board view that downstream modules can consume:

BOARD STATE

  • DIMENSIONS: rows: 8 / columns: 8
  • CELL REGIONS: logical grid cell regions
  • BOARD BOUNDS: detected board region
  • VALIDATION STATE: accepted / rejected candidate state

08 / TECH STACK

TECH STACK

  • Python
  • OpenCV
  • Computer Vision
  • Image Processing
  • Automation
  • BlueStacks
  • Windows

09 / STATUS

PROJECT STATE

SYSTEM ID PUZZLEMIND-S1

STATE ACTIVE R&D

DEPTH COMPUTER VISION / SCREEN UNDERSTANDING

Vision layer focused on robust board localization, grid validation and structured screen-state extraction.