facebookresearch / habitat-sim

A flexible, high-performance 3D simulator for Embodied AI research.
https://aihabitat.org/
MIT License
2.48k stars 406 forks source link

Dataset Download Issue: No train.json.gz and val.json.gz files #2357

Closed RV2005 closed 3 months ago

RV2005 commented 3 months ago

Habitat-Sim version: 0.3.1

I am trying to reproduce the results of a research paper, and trying to download the dataset required, using the following line: python -m habitat_sim.utils.datasets_download --username <api-token-id> --password <api-token-secret> --uids hm3d

After downloading, the data is not in the required directory format. Specifically, the train/ and val/ folders are like this:

scene-dataset/hm3d/
  val/ 
    ....
    00847-bCPU9suPUw9/  
    00898-8CRYizAb6yd/
    hm3d_annotated_val_basis.scene_dataset_config.json

same for train/

According to the description and what I've read so far, I was expecting something like this:

val/
    content/
        ...
        00847-bCPU9suPUw9/ 
        00898-8CRYizAb6yd/

    val.json.gz (This is missing completely! Likewise for train.json.gz)

Is there something wrong, or am I downloading the wrong thing? Apologies if this is trivial, I'm new to habitat.

Thanks in advance!

RV2005 commented 3 months ago

Found the answer! The train/val/minival datasets only contain the glbs and semantic information, and aren't supposed to have anything else. For the expected files, I had to download the Objectnav dataset from this link This contains files like val.json.gz and the required directories which were needed for the paper I was trying to reproduce.

Seems like a trivial case, but the paper didn't specify the required downloads :(