If you only want to upload your results to KITTI, you do not need the code in this repository. Just follow the instructions in the official KITTI site. This fork is useful if you want to evaluate your results using a custom validation set before uploading the final results to the public benchmark.
The evaluation is independent of the algorithm/framework that you are trying to evaluate. Just generate result files according to the KITTI format.
As I said in the previous point, this evaluation script is not aware of the detection method itself, so you need to measure detection time beforehand.
I recommend you to get familiar with the KITTI object dataset before trying the evaluation scripts.
Hi,I want to upload my results to Kitti. How do I use evaluate_object.cpp under caffe?How to test speed?