Data processing

Data processing is a fundamental step in converting event-based data into a format that is compatible with machine learning frameworks.

In this section, we will see:

• How to pre-process data to create tensors. Our training loops and our inference pipelines use tensors as inputs. This section will show you how to pre-process event-based data into a format that is compatible with PyTorch.

• How to convert a dataset into HDF5 tensor files compatible with the Metavision ML training loops. Training loops require a dataset in HDF5 tensor format, which is both smaller and faster than DAT or RAW files for Machine Learning.