Hi,I want to ask if the overall testing process is correct：I manually divide the data set into 3:7 randomly, run test.py and take 70% of the test set as input to get 10 txt test result files. The content of each txt file includes such as: 0 pedestrian 12 82 0.5092090964317322Posted by: Rainbown2021 @ March 18, 2021, 7:22 a.m.
You do not need to split the testing data. Please just submit all the results (10 TXTs), the evaluation server will split for you.
As for TXT format, note that you need to transform the range and angle to meters and radius, instead of the pixel grid. Please refer to the tutorial (https://github.com/yizhou-wang/cruw-devkit/blob/master/tutorials/cruw_devkit_tutorial_rod2021.ipynb) for how to do the transformation.
Can you provide your complete test set including pictures and tags? There are also codes for verifying the results. In this case, although the competition is over, I can still verify the performance of my optimized model。Thank you。Posted by: Rainbown2021 @ April 2, 2021, 6:16 a.m.