The goal of this project is to analyse and accelerate the task of predicting the number on an MNIST image using a pre-trained neural network. We would analyse the case of performing the entire computation on software in the PicoRV32 processor to identify the primary bottlenecks. Based on this analysis, an accelerator would be designed, analysed and interfaced, aiming to improve overall performance.
The baseline code for the project is available here
The accelerator designed is available here
The implementation with the accelerator integrated with the PicoRV32 is available here
The link to Video demonstration of project here
- Sai Gautham Ravipati, EE19B053
- Shashank Nag, EE19B118
- Vishnu Varma V, EE19B059