Improving the Accuracy of AI-Generated Images

TUMO Labs

Horizontal Line

As part of its Project-Based learning Program, TUMO Labs is launching a new initiative focused on enhancing the accuracy of images generated by artificial intelligence using only technical specifications. In the initial phase, the project will focus on accurately rendering lighting devices: lamps, chandeliers, and other fixtures.

The program participants will receive a scholarship.

Project Goal

The end goal is to develop a tool that improves the quality of interior design workflows.

Expected Results

Students will build a system that:

Accurately generates images of various lighting fixtures
Automates the process of generating technical specifications
Helps compare and refine AI-generated visuals

Students Will Learn To

Work with NLP models
Understand image generation processes
Use PostgreSQL/MySQL, React, and Google Cloud
Work with a mix of cutting-edge open-source and enterprise models for image generation. These include OpenAI’s DALL·E 3, Google DeepMind’s Imagen 2, Stability AI’s Stable Diffusion XL, and Amazon Bedrock’s Titan Image Generator.

Who Can Apply

To participate in the program, applicants should:

Have knowledge of mathematical analysis, linear algebra, probability, and statistics
Familiarity with discrete mathematics and graph theory is a plus
Be proficient in Python, including the following libraries:
- numpy, pandas – for data manipulation
- matplotlib, seaborn, or plotly – for data visualization
- scikit-learn – for data preprocessing, evaluation metrics, and classical machine learning tensorflow – նեյրոնային ցանցեր կառուցելու և մարզելու համար
- torch or tensorflow – for building and training neural networks
Be comfortable using Jupyter Notebooks or Google Colab for rapid prototyping and sharing ideas
Be familiar with Git and GitHub for version control, pull requests, and collaborative code review

Program Timeline

Duration: 3 months with a possible 1-month extension
Application deadline: September 12
Program start date: September 17

APPLY