Examining Parkinson’s symptoms is labor-intensive and constrained by a shortage of neurologists in underserved regions. In this thesis, you will generate synthetic training data by transforming low-poly 3D models of hands and feet into photorealistic images with labeled key points. Using techniques like pix2pix GANs, you’ll explore and optimize the process to create a diverse and enriched dataset, enhancing pose estimation models for AI-driven Parkinson’s symptom analysis.

💡Areas of Interest: Generative AI, Controllable Image Generation, Few-shot Learning

This master’s thesis is part of the graduation project ‘Ontzorg de zorg, zorg voor jezelf!’. This project gives the healthcare sector a digital boost through automation and data analysis, allowing caregivers to spend more time with patients while enabling patients to take control of their personal health data.

The examination of Parkinson’s symptoms is highly labor-intensive, as it requires multiple trained neurologists to thoroughly analyze hand and leg movements. A group of hospitals aims to extend Parkinson’s treatment in the Netherlands to regions where such care is currently unavailable due to a shortage of trained neurologists. To achieve this, they want to make a pre-selection through the help of AI on a smartphone. Part of the solution is a pose estimation model to track hand and foot movements. However, the current dataset for training the model is neither large nor diverse enough. The goal is to enrich the dataset with highly varied data while minimizing the need for manual effort.

Important features for detecting hands and feet could be better balanced by enriching the training dataset with synthetic data. To prevent manual labeling of the training data, it would be ideal if the ground truth for tracking key points is synthesized along with the training images. By converting low-poly 3D models with key point rigs into photorealistic images, a synthetic dataset of labeled training images could be generated. Your task will be to investigate how to transform low-poly models into photorealistic images using AI.

The Assignment

You will create a training dataset consisting of low-poly 3D body rigs, focusing on the hands and feet. Next, you will explore transforming these low-poly models into photorealistic images using a pix2pix GAN model. You may also fine-tune the model to optimize hyperparameters for generating the best results. Your findings will be documented in a thesis and presented during your thesis defense.

About Info Support Research Center

We anticipate on upcoming and future challenges and ensures our engineers develop cutting-edge solutions based on the latest scientific insights. Our research community proactively tackles emerging technologies. We do this in cooperation with renowned scientists, making sure that research teams are positioned and embedded throughout our organisation and our community, so that their insights are directly applied to our business. We truly believe in sharing knowledge, so we want to do this without any restrictions.

Read more about Info Support Research here.

About Info Support

Info Support specializes in custom software, data/AI solutions, management, and training and is active in the Finance, Industry, Agriculture, Food & Retail, Mobility & Public, and Healthcare sectors. We provide solid and innovative solutions for complex and critical software issues. Our headquarters are located in Veenendaal (NL) and Mechelen (BE). At present, approximately 500 employees are employed by Info Support.

Info Support's working method is characterized by a number of core values: solidity, integrity, craftsmanship, and passion. These core values are intertwined in our work and the way we interact with each other.

To ensure that all employees are always up to date with the latest developments, Info Support has an in-house knowledge center that eagerly satisfies the hunger for more or different knowledge and skills.

B2 language proficiency in Dutch is required.