VictorIA

A modular Python playground where game strategy, image annotation, and robotics co-exist inside one workflow.

Game AI Computer Vision Robotics
Role Technical Lead · Systems Designer
Timeline 2023 – 2024
Stack Python · OpenCV · NumPy
Status Research Playground
VictorIA robot arm positioned beside a physical Connect 4 board

Demo video

VictorIA in motion

Experiment lanes

What VictorIA juggles on a given night

  • Strategy lab

    Connect 4 engines battle through scripted tournaments while logging heuristics that other modules can steal.

  • Vision studio

    Annotation workflows mark cells, tokens, and defects with the same tooling, producing training data for future models.

  • Robot pit

    Serial adapters drive a tabletop arm through rehearsed motions that replay strategy or vision outputs safely.

Shared backbone

How one codebase serves three disciplines

Connect 4 brain

Game engine simulates futures and exposes APIs that other experiments reuse for decision loops.

  • Search depth adjusts based on available compute and latency budgets.
  • Board telemetry feeds into robotics pick-and-place routines.

Vision lab toolkit

Annotation loop writes normalized coordinates and masks to Processed_Cells/ for model training.

  • Shared schema lets robotics modules simulate cell pick-up sequences.
  • Plugins support new imaging modalities without rewriting UI.

Robot testbed

Unified adapter translates strategy outputs or vision detections into serial commands.

  • Simulation mode replays experiments for analysis without hardware.
  • Safety interlocks ensure motions stay within calibrated bounds.

Ideas travel between experiments

A better heuristic in Connect 4 informs the robot pick-and-place strategy; improved annotation flows sharpen the perception stack. VictorIA keeps all disciplines speaking the same language.

Flow

How the platform runs a scenario

01
Mode Selection

Launch main.py and pick a lane—game AI duel, image point lab, or robot exercise.

02
Data Capture

Load boards, ingest imagery, or read hardware sensors; normalize everything through utils/ adapters.

03
Analysis & Decision

Strategy modules, detectors, or heuristics transform the state into actionable commands or annotations.

04
Actuation & Export

Robotics routines move pieces, processed images land in Processed_Cells/, and logs capture performance metrics.

Gallery

Artifacts from experiments