Testing with test set

[holtvogt]

My custom dataset is currently organized in a KITTI format:

mmdetection3d
├── mmdet3d
├── tools
├── configs
├── data
│   ├── custom
│   │   ├── ImageSets
│   │   ├── testing
│   │   │   ├── velodyne
│   │   ├── training
│   │   │   ├── label
│   │   │   ├── velodyne

As mentioned in #1691, the testing config is, e.g. in KITTI dataset, the same as the validation config. When testing with tools/test.py, we are basically testing the model on the validation set which already has been used to fine tune the model during the training process, am I right? Just replacing the split and ann_file property with the corresponding testing name doesn't do the trick here as it can't find the annotations. Do we need to add another label folder inside the testing folder or how can we achieve to test on the real test set which the model hasn't seen before?