Back to projectsProject PDF
Machine learning experiment

Animatic

An experimental pipeline that takes a front-facing 2D character image and returns a simple textured mesh with an armature.

Academic researchResearch projectPreact · Flask · TensorFlow · Python
My contribution

Development of the image upload, model integration, and browser preview workflow for a research project.

Research preview

The public interface is available; model generation depends on its separate backend.

Interface preview01 / 01
Animatic interface

A thesis-era experiment with input constraints. It builds a simple textured mesh, not a full 3D reconstruction. Demo availability depends on its hosted model backend.

Animatic interface
The Animatic character image upload interface

The challenge

Getting from an image to something that can be inspected in a 3D viewer involves several separate steps. This project explores connecting those steps in one browser flow.

What I built

  • An image upload flow connected to a machine learning backend.
  • 2D keypoint prediction followed by simple mesh and armature generation.
  • GLB output, browser preview, and recent-file storage with IndexedDB.

Key implementation decisions

Keep the experiment’s limits clear

The pipeline targets a suitable front-facing character image and creates a simple textured mesh with an armature.

Make the output inspectable

A browser preview and GLB output provide a direct way to inspect the result; recent files are stored locally with IndexedDB.

The result

An end-to-end experiment that produces a rigged GLB from a suitable character image.