Job Title Master Thesis – Designing Ternary Neural Networks for Edge Vision Job Description Category Machine Learning, Computer Vision, Embedded Systems, Computer Science Scope 2 Master students, 30 hp each Background Deep-learning models have transformed computer vision, but deploying them on edge devices can be constrained by memory use, power consumption, and available compute. These limits matter in cameras and other devices that need to analyze images continuously while leaving resources available for other tasks. One possible way to reduce a model’s resource requirements is to constrain its weights to three values: −1, 0, and +1. This is called a ternary neural network (TNN). Ternary weights can reduce storage and memory traffic, and may allow some computations to use additions, subtractions, or skipped contributions instead of general multiplications. Whether this results in a faster or more energy-efficient system depends on the model, implementation, and target processor, so the trade-offs need to be measured. This thesis will investigate how to train and design TNNs for an edge-vision task. Starting with existing object-detection models, the students will assess the impact of ternary weights and explore ways to adapt the architecture and training process to retain useful accuracy. Possible areas of investigation include network blocks, quantization-aware training methods, activation precision, and mixed-precision designs. The resulting models will be benchmarked against suitable higher-precision or INT8 baselines. Evaluation may include accuracy, model size, CPU latency, throughput, and—where measurement equipment is available—energy use. The immediate focus is software running on available edge hardware. In the longer term, compact low-precision networks may also be relevant to specialized compute-in-memory hardware, such as memristor crossbars. Working with this type of hardware is not required for the thesis. Goals Investigate methods for training and designing ternary neural networks for edge-vision tasks. Assess the impact of ternary weights on existing object-detection models. Explore ways to adapt network architectures and training methods to retain useful accuracy. Compare approaches such as different network blocks, quantization-aware training, activation precisions, and mixed-precision designs. Benchmark TNNs against suitable higher-precision or INT8 baselines. Evaluate the trade-offs between accuracy, model size, CPU latency, throughput, and—where measurement equipment is available—energy use. Identify model and implementation choices that can provide a meaningful advantage on real edge hardware. Consider the potential relevance of ternary models to future compute-in-memory hardware. Who are you? For this thesis proposal, we are looking for two students with a strong interest and curiosity in deep learning, computer vision, model compression, embedded systems, or computer architecture. You are studying a Master’s program in Computer Science, Electrical Engineering, Computer Engineering, or a related field. You have some programming skills, preferably in Python, and experience with PyTorch or a similar deep-learning framework. OK, I am interested! What do I do now? You are valuable to us – how nice that you are interested in one of our proposals! There are a few things for you to keep in mind when applying. Applications are accepted in both Swedish and English, and you apply via the proposal advert. The announced thesis is open only to students affiliated with a Swedish University/College, either directly or via an exchange program. When the thesis proposal states that it includes two students working together, we would like you to apply in pairs. In these cases, send one application each but make sure to clearly state in your application who your co-applicant is. If you have any questions regarding this, please do not hesitate to contact us. Please attach your CV, cover letter, and University/College grade summary. Who to contact for any questions regarding the position! Magnus Svensson, Engineering Manager, [email protected], +46 46 2879187 Type of Employment Thesis Worker (Fixed Term) Posting End Date 2026-11-27 Certain roles at Axis require background checks, which means applicable verifications will be done in these recruitments. Notice will be provided before we take any action. About Axis Communications We enable a smarter and safer world by creating innovative solutions for improving security and business performance. As a network technology company and industry leader, we offer solutions in video surveillance, access control, intercom, and audio systems, enhanced by intelligent analytics applications. With around 6,000 committed employees in over 50 countries, we collaborate with partners worldwide. Together, we thrive in our friendly, open, and collaborative culture and inspire each other to think beyond the expected. United by our commitment to inclusion, diversity, and sustainability, we consistently seek to develop our skills and way of working. Let´s create a smarter and safer world For more information about Axis, please visit our website www.axis.com. Listen to Axis Communications Playlist We enable a smarter, safer world by creating innovative solutions for improving security and business performance. As a network technology company and industry leader, we offer solutions in video surveillance, access control, intercom, and audio systems, enhanced by intelligent analytics applications. Axis for cities Axis podcast playlist

Master Thesis – Designing Ternary Neural Networks for Edge Vision.

more at Axis.

ASIC SoC System Architect, Lund

Sales Engineer - End Customer

Thesis – Exploring Wi-Fi Sensing for Future Intercom Applications

similar roles.

Senior ML Engineer | Kimchi (Inference Stack Optimization)

Senior Data Scientist

Software Engineer — AI-Native Full Stack

Research Operations & AI Enablement, Winter 2027 Co-op

Machine Learning Engineer Co-op

Case Manager