search

jzhzhang / 3DAwareNav

[CVPR 2023] We propose a framework for the challenging 3D-aware ObjectNav based on two straightforward sub-policies. The two sub-polices, namely corner-guided exploration policy and category-aware identification policy, simultaneously perform by utilizing online fused 3D points as observation.
MIT License
57 stars 3 forks source link

Dataset #1

Closed YYDS-cc closed 1 year ago

YYDS-cc commented 1 year ago

Excellent work！ This is my first time working with datasets for object navigation and I have a few questions to ask. According to the readme, I am downloading the habitat zip from the (MP3D) dataset, how much storage space does this require, and furthermore, how do I split the downloaded dataset into the form of a readme?

jzhzhang commented 1 year ago

Thank you for your interest in our work! I believe the scene dataset may require approximately 30-40GB of storage (the episode dataset won't occupy much space). Additionally, you don't need to be worried about the training and validation data as it has already been separated in the episode dataset.

YYDS-cc commented 1 year ago

Thanks for the reply, I have downloaded and unzipped habitat.zip, which has 90 scenes（20.8GB）, but I didn't find any information about the episode dataset in the official download file. How can I download it?

jzhzhang commented 1 year ago

The details can be found in habitat-lab dataset, and the you should download the Matterport3D objectnav_mp3d_v1.zip

YYDS-cc commented 1 year ago

Thank you so much!