search

tue-mps-edu / asd-engd-project-2019-thermal-object-detection

This is the repository for the ASD PDEng module 2 project (2020) for AIIM
BSD 3-Clause "New" or "Revised" License
1 stars 0 forks source link

Issue 67 68 opt #71

Closed Hrayo712 closed 4 years ago

Hrayo712 commented 4 years ago

This Pull Request includes the necessary changes to solve Issue 67 and Issue 68. It implements the necessary code to run online inference on a non-optimized graph, as well as an optimized TensorRT Engine (Keeping the previous functionality). Also, the FPS measurement is implemented by storing all the FPS data to an empty CSV file, which can be analyzed afterwards to obtain the Average FPS.

Additionally, minor changes and fixes to README.md files on src/ and the main README are included.

IMPORTANT: As it is my Issue, testing should be done by me. However, due to the current situation with the home-office, it is impossible for me to run this testing, as the target hardware (Jetson Xavier) is not available to me. Therefore, It is necessary for @koenelands to perform Regression testing as part of the review process of this Pull Request. For this the following has to be verified:

  1. The code is able to run online inference using the non-optimized frozen graph (Tensorflow's .pb) on video Feed from the Thermal Camera while recording FPS on the CSV
  2. The code is able to run online inference using the optimized TensorRT engine on video Feed from the Thermal camera while recording FPS on the CSV
  3. The code is able to run online inference using the optimized TensorRT engine on video Feed from the Thermal camera without recording FPS. This is to guarantee the previous functionality is intact, ensuring the new features do not impact the previous ones.

closes #67 and closes #68

koenelands commented 4 years ago

Looks good. It works on the Jetson. I made .csv files and will add it in the PR for #73

Hrayo712 commented 4 years ago

Regression was tested and functionality was validated.