The EUPEX researchers share their results and lessons learnt!
We are organising monthly webinars about the outcomes of EUPEX (full programme here). Our second webinar will be dedicated to one of the many applications examined or optimized under EUPEX.
The presentation will introduce an AI framework developed at the University of Zagreb designed to parse neural networks in ONNX format efficiently. By eliminating the reliance on external libraries, it simplifies the optimization of the necessary operators. This approach not only enhances the interpretability but also provides users with the flexibility to customize the code for diverse architectures. While the framework can be applied to various AI tasks, it has been integrated into Bolt65 for a real-life use case: vehicle detection. By demonstrating its effectiveness in this domain, we aim to highlight its potential for future AI applications using different vectorization approaches and across different architectures.
Speaker: Hana Ivandić, Research and Teaching Assistant & PhD Student at Faculty of Electrical Engineering and Computing, University of Zagreb