How to initialize environment outside of main repo? about habitat-lab HOT 4 CLOSED

The "just make it work" fix is to do ln -s /path/to/habitat-api/data while in the folder for the different project's repo

from habitat-lab.

After changing the DATA_PATH to the absolute path of the config, I get:

~/svl/habitat/feature_selector/habitat-sim/habitat_sim/simulator.py in _config_backend(self, config)
     52     def _config_backend(self, config: Configuration):
     53         if self._sim is None:
---> 54             self._sim = hsim.SimulatorBackend(config.sim_cfg)
     55         else:
     56             self._sim.reconfigure(config.sim_cfg)

ValueError: Cannot load: data/scene_datasets/gibson/Adrian.glb

from habitat-lab.

You can create a new train pointnav dataset setup, the new setup will be mostly identical to the current setup, you can use the current pointnav_gibson.zip as the starting point. The train dataset for Gibson is split over different houses, so the folder pointnav/gibson/v1/train/content contains files like Adrian.json.gz, Albertville.json.gz which contains the different episodes inside the respective houses. Each episode contains a scene_id which is of the type data/scene_datasets/gibson/Adrian.glb. You will have to modify this value to the location where you have stored the glbs.

Let me know if you have anymore questions and thanks for trying this out, I do recognize that this kind of setup is slightly convoluted and we are trying to simplify this setup for future.

from habitat-lab.

@bradleyemi with #77 we have parameter config.DATASET.SCENES_DIR to control the location of scenes dataset and get rid of data symlink requirement. Let us know if you have any questions regarding that and feel free to reopen the issue.

from habitat-lab.