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.
Semi-supervised learning (SSL) is a machine learning paradigm that leverages both labelled and unlabeled data to improve model performance. The traditional SSL approach involves training models with pseudo-labels assigned to unlabeled data. However, this technique often encounters challenges in situations where the pseudo-labels are noisy, leading to overconfident predictions and suboptimal performance, particularly in complex tasks. To address these shortcomings, recent advances in SSL have focused on regularization methods to improve model robustness and generalization. Among these techniques, one stands out: consistency regularization, which emphasizes the importance of maintaining consistency between predictions of the same image when it is subjected to various augmentations. This consistency across different augmentations helps models learn more invariant representations and can effectively reduce overfitting on noisy pseudo-labels.
- Method Development: The primary goal of this thesis is to advance the field of semi-supervised learning by proposing and developing a novel framework for SSL that integrates the concept of multi-view consistency. This framework aims to address limitations inherent in traditional SSL methods, offering a robust approach to harnessing unlabeled data effectively across diverse machine-learning applications.
- Performance Enhancement: Central to this research is the objective to design SSL algorithms that exhibit heightened resilience to the challenges presented by noisy pseudo-labels. The research aims to quantify the efficacy of the novel framework in strengthening model robustness and its capability to tackle intricate tasks.
- Real-World Applications: Beyond theoretical contributions, the research seeks to validate the practical utility of the proposed framework. This involves applying the framework to real-world scenarios across various domains. The goal is to demonstrate that the framework has the potential to significantly elevate the performance of SSL models in a wide array of real-world applications.
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 image classification, object detection and semantic segmentation is highly valued, it is not mandatory, as this thesis will provide the necessary context and guidance for applying semi-supervised learning to these tasks.
Required application material
- Cover letter
- Transcripts (with grades)
All application material must be in English.
Work starts in beginning of January 2024
Interested in joining Univrses as our new Master Thesis student? Submit your application - All application material must be in English!