Chalmers Open Digital Repository
Välkommen till Chalmers öppna digitala arkiv!
Här hittar du:
- Studentarbeten utgivna på lärosätet, såväl kandidatarbeten som examensarbeten på grund- och masternivå
- Digitala specialsamlingar, som t ex Chalmers modellkammare
- Utvalda projektrapporter
Forskningspublikationer, rapporter och avhandlingar hittar du i research.chalmers.se
Enheter i Chalmers ODR
Välj en enhet för att se alla samlingar.
Senast publicerade
- Vision Language Model Based Systems for Optical Character Recognition of Historical Swedish Newspaper Material(2026) Johansson, Martin; Waginder, SelmaThe ability to digitize old scanned newspapers plays an important role in improving searchability and making information accessible. To convert text in images into machine-readable form, an Optical Character Recognition engine is employed. In this thesis, a dataset Swedish newspaper material from 1818-1904 is used. The report investigates whether small to medium-sized open-source Vision Language Models are competitive for Optical Character Recognition compared to traditional models. It is found that a fine-tuned Qwen3-VL-8B-Instruct in combination with a simple repetition trimmer is able to outperform the traditional OCR engine Abbyy FineReader version 11.1.16 by 68.5% in terms of CER on this particular dataset and set a new record at 1.930% CER. This thesis demonstrates that the current generation of small open source Vision Language Models are highly competitive with traditional OCR engies for transcription of 19th century Swedish newspaper material. The thesis also thoroughly investigates the particular quirks and failure modes of different OCR systems through a qualitative analysis. Our best model performs no better on the training set than on the test set, suggesting that our finetuning was bottle necked by the LoRA adapter size and that one could potentially achieve an even stronger model with a larger adapter.
- Protecting Patient Privacy in Healthcare Analytics with Fully Homomorphic Encryption and Differential Privacy(2026) Wennerberg, Joakim; Isgandarli, ShahnurData analytics in the healthcare domain requires access to sensitive patient information, creating conflicting interests between the need for usability and privacy requirements. Fully Homomorphic Encryption (FHE) enables computation on encrypted data, while Differential Privacy (DP) protects individuals against inference attacks by introducing controlled noise into aggregate query results. This thesis investigates the combined use of FHE and DP for privacy-preserving healthcare analytics and evaluates the resulting privacy guarantees, performance, and practical limitations. Three aggregation queries are implemented and evaluated in a multi-party privacy-preserving system using multiple FHE schemes and libraries, including the BFV, BGV, CKKS, and TFHE schemes using the Microsoft SEAL and Concrete FHE libraries. Performance, accuracy, ciphertext expansion, and compliance to confidentiality and availability requirements are assessed using a synthetic healthcare dataset. The results show that combining FHE and DP strengthens protection against eavesdropping and membership inference attacks compared to using either of the methods alone. However, the increased privacy comes at a large cost in performance and usability. Encrypted query execution is orders of magnitude slower than plaintext execution, and current FHE libraries provide limited support for common statistical operations. Additionally, even in the best case, ciphertexts span several megabytes for a single value, although this can be partially mitigated through compression prior to storage or network transmission. Finally, executing homomorphic computations in an insecure environment could expose encrypted data to side-channel attacks such as power measurement or timing attacks. These limitations represent a significant hindrance for large-scale deployment of FHE in statistical contexts. Improvements to the FHE ecosystem in regards to performance and usability could enable future large-scale deployment.
- Distributed Sketching Pipelines for Data Mining and Analytics(2026) Mentzer, Jonatan; Bruhn, GustavThe rapid growth of data generated by modern industrial systems at the edge poses significant challenges for efficient data management. Processing the data directly on the edge device offers benefits, including improved throughput, scalability, and reduced bandwidth usage. Another effective approach is summarizing the data using sketches. Sketches are stochastic data structures that can greatly compress data while preserving essential statistical properties. This thesis investigates the conditions under which it is beneficial to offload sketching computations to edge devices. The study evaluates the throughput and latency of two systems across multiple configurations, designed to reflect real world scenarios, comparing a federated architecture with a centralized architecture. The results indicate that, across all evaluated scenarios, executing computations at the edge increases the maximum throughput, especially when the number of edge devices increases. The thesis explores the trade-offs between scalability (in form of throughput with increasing set of vehicles) and data freshness (in form of latency due to the micro-batch sizes, i.e. the frequency of data summarization).
- Target-Guided Trajectory Generation for Controllable Traffic Scenarios - A novel conditioning method for diffusion-based trajectory generation enabling controllable traffic scenario synthesis(2026) Bredin, Jacob; Haraldsson, LinusAutonomous vehicles (AVs) must be evaluated under rare and hazardous driving conditions to ensure safety and reliability. However, creating such safety-critical scenarios is made difficult for several reasons. They occur infrequently in real-world data and are costly to reproduce through physical testing, while existing simulation methods often yield unrealistic behaviors. This thesis explores generative modeling as a tool for producing realistic and controllable scenarios for closed-loop evaluation of AV systems. We introduce a novel diffusion-based method for generating adversarial trajectories, with a focus on Classifier-Free Guidance (CFG) to steer agents toward defined targets. The approach incorporates target information during training, uses data augmentation to improve robustness, and applies trajectory optimization to enhance accuracy. Building on the Versatile Behavior Diffusion (VBD) framework, our method strengthens controllability while preserving realistic motion patterns. The experimental results show that CFG improves guidance performance without any additional computational cost during inference, which has been a major limitation of prior approaches, while still matching the accuracy of classifier-based guidance. When combined with classifier-based guidance, CFG yields substantial improvements in target accuracy and reduces the number of required guidance iterations. Furthermore, direct trajectory optimization is shown to further refine target accuracy, although it introduces trade-offs with respect to adherence to traffic regulations. Collectively, these findings establish an efficient and versatile framework for the generation of safety-critical driving scenarios, thereby advancing the methodological foundation for rigorous evaluation of autonomous vehicle systems.