Inference

In this section, we will see how to use our trained models. You can use either one of our pre-trained models or train a model yourself.

Using a model at inference-time typically consists of a few standard steps:

• The setup:

  • Load the model

  • Load a data file or open an event-based camera

  • Start streaming

• The processing loop:

  • Gather data from an event-based camera or from a RAW or DAT file

  • Pre-process the data using the same format used for training the model

  • Perform inference by passing the preprocessed data to the model

  • Gather and process the results

More details on the inference loop can be found in these examples:

• Detection and tracking pipeline

• Use directly a torchjit model