Tesi etd-11112025-171045 |
Link copiato negli appunti
Tipo di tesi
Tesi di laurea magistrale
Autore
BALDACCINI, MARCO
URN
etd-11112025-171045
Titolo
Design, Implementation, and Verification of a ResNet18-Like Convolutional Neural Network exploiting the FPG-AI Platform
Dipartimento
INGEGNERIA DELL'INFORMAZIONE
Corso di studi
INGEGNERIA ELETTRONICA
Relatori
relatore Prof. Fanucci, Luca
correlatore Ing. Nannipieri, Pietro
supervisore Dott. Bocchi Tommaso
correlatore Ing. Nannipieri, Pietro
supervisore Dott. Bocchi Tommaso
Parole chiave
- AI
- CNN
- FPG-AI
- FPGA
- neural network
- ResNet
Data inizio appello
05/12/2025
Consultabilità
Non consultabile
Data di rilascio
05/12/2095
Riassunto
This thesis presents an implementation of residual networks in the FPG-AI toolflow. The toolflow scripts were modified to enable execution of residual layers, and the HDL description of the architecture was changed to implement skip connections. Several Python scripts were developed to automate the generation of hardware parameters required by the tool to support this new type of network. A ResNet18–like network was implemented in Python and accelerated using the FPG-AI toolflow. The results of this acceleration on a space-grade and a commercial-grade FPGA board are compared with the state of the art in residual implementations.
File
| Nome file | Dimensione |
|---|---|
Tesi non consultabile. |
|