Master thesis: Semantic Communications with Foundation Models for Vehicle and Road User Detection
Core
Develop a semantic communication system to transmit task-relevant spatial features from vision foundation models (CLIP, DINOv2) over noisy wireless channels for vehicle and road user detection.
Role type
Master's thesis researcher (Semantic Communications & Computer Vision)
Builds
A semantic-communication system with a learned bottleneck, quantization, and lightweight decoder for object detection
Domain
Telecommunications (5G/6G) and Computer Vision
Deliverable
production ML models
Required skills
Python, PyTorch, Deep Learning (CNNs, Transformers), Wireless Communications (AWGN, SNR), Object Detection
Preferred skills
Familiarity with CLIP, ViT, DINOv2, YOLO, Faster R-CNN, FCOS
Technologies
CLIP-ViT-B/16, DINOv2-ViT-B/14, COCO dataset, PyTorch, AWGN simulation
Responsibilities
Implement learned bottleneck and decoder, train end-to-end with noise injection, compare foundation models against task-specific baselines, analyze mAP vs SNR
Seniority
Master's student researcher
