Back to projects

Computer vision / Software & Backend

HANDORA

Experimental computer-control prototype using MediaPipe/TFLite hand tracking, with the left hand controlling the cursor and the right hand handling clicks and scrolling.

Prototype
  • MediaPipe
  • TensorFlow Lite
  • Computer Vision
  • Hand Tracking
  • Gesture Interaction

Original code-generated category illustration; not a product screenshot.

Overview

HANDORA is my independent prototype exploring computer input through camera-based hand tracking. The MediaPipe/TFLite approach assigns different roles to each hand: the left hand controls the cursor, while the right hand handles clicking and scrolling. Hand detection and control stability remain development limitations. The work is distinct from the earlier media-control prototype.

Problem and objectives

Camera-based touchless computer control needs to distinguish cursor movement, clicking and scrolling. Losing a hand track or interpreting unintended motion as a command can make the interaction unstable.

My role

Prototype development combining MediaPipe/TFLite hand tracking with two-handed interaction: left-hand cursor control and right-hand clicking/scrolling.

Solution approach

The prototype uses a MediaPipe/TFLite-based hand-tracking approach. Left-hand movements map to cursor control, while right-hand movements handle clicking and scrolling. This separates gesture interpretation from computer-input generation. Detection and stability issues remain open parts of the prototype rather than solved capabilities.

Architecture flow

01

Camera image input

02

MediaPipe/TFLite-based hand tracking

03

Left hand: cursor control / right hand: clicking and scrolling

04

Conversion of interpreted gestures into computer-control commands

Conceptual flow derived from project descriptions; not an application screenshot.

Core components

  • Camera-based hand tracking.
  • Cursor control with the left hand.
  • Clicking and scrolling with the right hand.
  • A prototype assigning different control roles to the two hands.

Technical challenges

  • Maintaining stable control when hand detection or tracking is lost.
  • Cursor jitter, gesture stability and unintended commands.
  • Effects of camera conditions, lighting and hand visibility on interaction.

Engineering decisions

  • Assign cursor movement and clicking/scrolling to different hands.
  • Maintain a separate scope from the earlier gesture-media-control project.
  • Keep detection and stability problems visible as development limitations.

Testing and validation

  • This technical account follows the prototype approach described by the project owner.
  • No separate validation report establishes measured detection accuracy, latency or device compatibility.

Outcomes

  • An experimental control prototype separating left-hand cursor movement from right-hand clicking/scrolling.
  • The hand-tracking approach and remaining stability issues are documented.

Limitations and current status

  • Hand detection and control stability require further development.
  • Platform compatibility, accessibility and performance are not presented as validated outcomes.
  • No live demo or publicly accessible source-code link is available.

This account is based on information supplied by the project owner; it does not assert independently verified performance or publication outcomes.

Related work