According to the information disclosed on the competition website, the test environment seems to be unified, but this may not meet everyone’s needs. For example, I modified some of the code on mmdetection and need to be re-setup, but because the environment is fixed, my code cannot Run. Can I summit the result with a docker which configure the relevant operating environment？
Hi @VOVO, thanks for the question.
Like you mention, all of the official package versions are standardized in the runtime. But, you can include any custom code you write in your zip of your submission file and then import from that local code in your main.py.
For example, this is how we include some additional code in our benchmark: