Cogni-draw

AI-Driven 2D to 3D Sketching Application

Cogni-draw is an interactive prototype application that leverages AI techniques to convert 2D sketches into rudimentary 3D representations. By right-clicking and dragging over the framebuffer, users can capture their sketch to save them as PNG images to then be transformed into a 3D model, viewable in the model viewer dock.

---

Features

• Region Capture: Right-click and drag to select a rectangular region of the framebuffer, which is then saved as a PNG image.
• AI Export: ChatGPT api with 4o-mini and then dalle, working on segmentation api, might use meshy
• Note: I can download a model directly from meshy's website and this works, I have the api script for meshy, but need to buy tokens to use it. The demo down below is a proof of concept for now.

Prerequisites

• Just MSVC for now
• GLFW (for window and input management)
• GLAD or another OpenGL loader
• stb_image.h (for image loading)
• stb_image_write.h (for PNG output)
• stb_image_resize.h (for image resizing)

Demo I/O (stb save -> o4-mini vision -> dalle -> meshy(not included yet) -> OpenGL

2025-05-02.21-39-00.mp4

Building

1. Clone the repository:

   git clone https://github.com/anakin004/Cogni-draw.git
   cd Cogni-draw