Forums

ECCV 2022 Workshop SSLAD Track 5 - UNIFIED MODEL FOR MULTI-TASK LEARNING Forum

Go back to competition Back to thread list Post in this thread

> To the committee: Need more details of bev mask

We have been expolaring the AutoScenes dataset, and have the question below:
Is there any offcial instruction to decode bev mask from the dataset? We follow nuScenes / Cross View Transformers / BEVFusion for now, but not sure everything matches the challenge requirments, for instacne "resolution_meters" and "patch angle"

Posted by: terrywayne @ Aug. 3, 2022, 6:45 a.m.

parameter like 'resolution_meters' we follow the default setting in nuscenes-devkit. 'patch_angle' is not set but calculated according to the ego pose calibration.

Posted by: SSLAD2022 @ Aug. 3, 2022, 9:37 a.m.
Post in this thread