Notice

Menghan Li is assembling a team of people who would like to work on this project. If you are interested, send her email: ml993@cornell.edu.

Make 3D

Client

Ashutosh Saxena, Assistant Professor, Computer Science Department
asaxena@cs.cornell.edu

Project

The Make3D software converts a single still image (e.g., one taken from a digital camera) into a 3D model. This 3D model can be viewed on a 3D display or a mobile device.

The code (90% in C++) currently runs on a desktop computer. See: http://make3d.cs.cornell.edu for some results, and contact the client for the newest version of the code.

The goal of this project is two-fold: (a) convert the current version of the software into production quality code, (b) rewrite parts of code, reorganize it so that it takes lesser amount of RAM and runs faster on a processor. The code will be written in openCV/C++, so experience with C++ is necessary for this project. The current version of code takes about a minute to run on a single core desktop computer, and takes about 700 MB of RAM. The goal is to have it take less than 300 MB RAM (with peak usage of 350 MB), and finish the processing in 10-15 seconds. This will allow the code to run on a cellphone, but restructuring the code, following good practices in software engineering, with no memory leaks, and well documented code is more important than the processing time. I.e., it is okay to have it take 30-40 seconds if other criterion are satisfied.

An optional goal is to write an Android app that converts the selected pictures on the cellphone gallery into 3D model on the device itself. The client writes that there is considerable demand for this application as there are many 3D display devices, but there is a lack of the ability to create 3D content. With ubiquitous smartphones, this could potentially become an extremely popular app.

It should be released on the Internet as an open source system with suitable licenses and provision for extensions.