search

VinAIResearch / ISBNet

ISBNet: a 3D Point Cloud Instance Segmentation Network with Instance-aware Sampling and Box-aware Dynamic Convolution (CVPR 2023)
Apache License 2.0
104 stars 22 forks source link

About scannetv2 hidden test upload #15

Closed 3171228612 closed 1 year ago

3171228612 commented 1 year ago

What else should be done to handle pred_masks after test output before submitting them to the authorities? After I output the masks, the assessment fails all the time

ngoductuanlhp commented 1 year ago

You should follow our instructions to export the predictions in the correct format of ScanNetV2 benchmark, i.e.: python3 tools/test.py configs/scannetv2/isbnet_benchmark_scannetv2.yaml <checkpoint_file> --out <output_dir>

The function save_single_instance in tools/test.py will map the predicted class_id to the corresponding benchmark_class_id, then write to the .txt file. Ensure that the total number of files in the test set is 100 and follow the instruction at https://kaldir.vc.in.tum.de/scannet_benchmark/documentation to zip the folder correctly.

3171228612 commented 1 year ago

Thank you for your explanation