On Tuesday, July 15, 2025, at 10:30 a.m., you are cordially invited to the final presentation of the Bachelor thesis by Mr. Ngapna Njikam Adam Mfoluop in seminar room of C3E (262A). The thesis is titled
“Evaluation and Analysis of Different Convolutional Neural Networks on a Massively-Parallel Vector Processor.”
Convolutional Neural Networks (CNNs) have automated object detection and image segmentation of large sensor data streams, especially in computer vision. Pursuing higher accuracy has led to more complex architectures and greater computational demands, requiring specialized hardware accelerators. In the BMBF-funded ZuSE-KI-AVF project at the Chair for Chip Design for Embedded Computing, a programmable vertical vector processor was developed for parallel computations. An accompanying converter framework imports CNN models, quantizes weights, and automatically generates the necessary memory-access and compute instructions. In this work, the framework was extended with multiple CNNs and thoroughly evaluated.