Investigate Synthetic Texture Generation Application for Integration into the Synthetic Data Pipeline

Applied Recognition Corp. (ARC), a Canadian software company based in Oakville, Ontario, develops facial recognition technology solutions designed to be both accurate and practical, enabling organizations to adopt biometric systems in a simple and cost-effective manner.ARC’s machine learning models depend on high-quality human facial image datasets to ensure reliable performance and accuracy in image detection and recognition. However, the limited availability of diverse, high-quality real-world facial data presents an ongoing challenge for training and testing purposes.ARC is collaborating with WIMTACH-CEDA to examine the technical limitations and operational challenges with generating synthetic face images using various artificial intelligence-based generation approaches.Building on previous research, this initiative focuses on identifying and documenting key system parameters and control features that influence consistency, scalability, and usability within the synthetic face image generation process.

This proposed project targets a critical operational challenge within ARC’s synthetic data pipeline: the need for scalable, consistent, and controllable texture generation methods. While the current pipeline supports variations in facial geometry and structure, limitations in texture consistency and bias control affect the reliability and repeatability of synthetic facial datasets at scale. Accordingly, the project aims to evaluate, optimize, and refine texture generation settings to enhance output consistency, reduce processing inefficiencies, and support large-scale data production aligned with ARC’s business and deployment requirements.The anticipated impact on ARC includes improved product scalability, accelerated model development, and enhanced revenue growth. By strengthening its synthetic data pipeline through improved texture control, ARC can efficiently expand training datasets and shorten development cycles, enabling faster deployment of facial recognition solutions. This approach reduces reliance on costly and sensitive real-world data, lowers compliance and acquisition costs, improves profit margins, and supports market expansion. The use of privacy-preserving synthetic data further strengthens customer confidence and reduces adoption barriers, contributing to sustainable long-term growth.

Faculty Supervisor:

Tenzin Jinpa

Student:

Partner:

Applied Recognition Corp

Discipline:

Computer science

Sector:

Information and cultural industries

University:

Centennial College of Applied Arts and Technology

Program: