Community Bonding: Week 3

1 minute read

Preliminaries

I identified some errors while using the existing frameworks. This week, we will continue to fix the issues found and discussed during the meeting. I also verified the inference time for PilotNet model (PyTorch implementation), the difference between the CPU vs GPU execution is (0.0265 vs 0.0192) seconds (mean inference time) for simulation of 1 minute.

When we save stats of a simulation a window shows all metrics in details. We will also use this in later phases to evaluate and compare our optimized models. I describe our new objectives next.

Objectives

Issues and Pull requests.

The execution

This week was dedicated to resolves the previously found issues. Moreover I also included new features which would be helpful in conducting experiments and evaluation. For the training part, I worked on finalizing installation packages, enable training scripts to use GPU, also update it to use new training datasets and evaluate on a validation set. Furthermore, the inference scripts on BehaviorMetrics were updated accordingly.

References

[1] https://github.com/JdeRobot/BehaviorMetrics
[2] https://github.com/JdeRobot/DeepLearningStudio