Basic setup for software model development only need conda(anaconda/miniconda) and jupyter notebook, where specific dependencied are all in environment.yaml file. Given that the size of the dataset, working at S3DF is preferable.
The Smart Pixel dataset is available at https://zenodo.org/records/10783560. There is also an available notebook for processing the data into .hdf5 file for easier processing.
Different model development notebook and existing models are all available at software directory.
Similarly, we have a environment file for setting up the conda environment. A setup.sh script is in HLS directory for you to set up license and GUI in rdsrv machine.
Source code for synthesis can by found under HLS/src, and all model synthesis part are store in <>onChip subdirectories. For CNN, since the RTL code are large, only non-automatic generated part are store in this repo.
The scripts are written specifically for myself, compatibilities are not promised and will be add later. Please feel free to contact Kenny Jia through email: hjia625@stanford.edu, or haoyi.jia@cern.ch. I would be happy to help with the debugging process or questions!