Do you want to create technology for the next generation of industry and society? Univrses is now looking for a Master Thesis student to join the team in Stockholm!

Univrses is a 3D Computer Vision and Machine Learning company based in Stockholm, creating high-end technologies for autonomous systems. We work in several different areas, but our main focus is on self-driving vehicles, mobile robotics and smart city development. Our team consists of hard-working and friendly people from all over the world with diverse backgrounds and expert knowledge in computer vision, robotics, machine learning, physics, math, software development and more.

Master Thesis at Univrses

Right now, we offer Master students the opportunity to be an integral part of the team while working on their Thesis. You will be working in a fun, stimulating and highly professional work environment together with our awesome team. This is truly a unique chance to work with some of the best scientists and engineers in Computer Vision and Robotics in the world.

Background

Semantic segmentation is a fundamental task in computer vision, playing a crucial role in various applications such as autonomous driving, medical image analysis, and object recognition. Deep learning models have shown remarkable performance in semantic segmentation, but their success largely depends on the availability of high-quality labeled datasets for training. The scarcity of diverse and comprehensive labeled datasets, especially in specific domains or for rare objects, hinders the development and evaluation of robust semantic segmentation models.

One promising solution to address the data scarcity problem is the generation of synthetic datasets. However, creating a synthetic dataset that accurately mimics real-world data remains a challenging task. Current synthetic datasets often suffer from unrealistic artifacts, which can lead to poor model generalization when trained on them.

Goals

The primary objectives of this research are as follows:

1. Diffusion Model Adaptation: Investigate the adaptation and extension of diffusion models for generating high-quality synthetic datasets for semantic segmentation. This includes developing methods to ensure that generated images exhibit realistic object shapes, textures, and environmental contexts.
2. Quality Assessment: Develop quantitative and qualitative metrics for assessing the quality of the synthetic dataset. Ensure that the dataset meets the necessary criteria for training and evaluating semantic segmentation models.
3. Performance Improvement: Evaluate the extent to which semantic segmentation models trained on the synthetic dataset outperform models trained on existing synthetic datasets or when real data is scarce.

Relevant material

• Denoising Diffusion Probabilistic Models
• Effective Data Augmentation With Diffusion Models
• DiffuMask: Synthesizing Images with Pixel-level Annotations for Semantic Segmentation Using Diffusion Models
• DatasetDM: Synthesizing Data with Perception Annotations Using Diffusion Models

Required background

Applicants are expected to possess a comprehensive understanding of deep learning and computer vision, which will be essential for engaging with the research effectively.

Proficiency in Python coding is a prerequisite for implementing and experimenting with the proposed techniques.

While prior experience with diffusion models is highly valued, it is not mandatory, as this thesis will provide the necessary context and guidance for their application in generating realistic synthetic datasets for semantic segmentation.

Required application material

• CV
• Cover letter
• Transcripts (with grades)

All application material must be in English.

Work starts in beginning of 2024

Next step

Interested in joining Univrses as our new Master Thesis student? Submit your application - All application material must be in English!