Semantic annotations not getting loaded

Hi, I am trying to visualize an ObjectNav agent but there seems to be an error in TopDownMap measure when it is trying to draw goals:

File ".../habitat-challenge-dev/habitat/tasks/nav/nav.py", line 784, in _draw_goals_aabb
    sem_scene.objects[object_id].id.split("_")[-1]
IndexError: list index out of range

Seems like semantic annotations are not getting loaded but I have the complete dataset.

[22:35:57:534470]:[Sim] Simulator.cpp(432)::instanceStageForSceneAttributes : Successfully loaded stage named : habitat-challenge-data/data/scene_datasets/hm3d/val/00839-zt1RVoi7PcG/zt1RVoi7PcG.basis.glb
[22:35:57:534517]:[Sim] Simulator.cpp(464)::instanceStageForSceneAttributes :
---
The active scene does not contain semantic annotations : activeSemanticSceneID_ = 0
---

Am I missing something here?

[srama2512]

@nnsriram97 - I was able to successfully evaluate the ddppo baseline checkpoint on HM3D minival with video saving. It is possible that the issue may be with the conda installation. We’re looking into it. In the meanwhile, here are the steps I took for evaluation (NOTE: I’m not loading semantic annotations here):

• Install challenge-2022 branch of habitat-lab

• Install challenge-2022 branch of habitat-sim from source (see here)

  python setup.py install --headless --with-cuda
  

• Make sure that HM3D scenes, semantics, and ObjectNav datasets are downloaded in <habitat-lab-path>/data

• Download ddppo checkpoint

cd <habitat-lab-path>
mkdir -p experiments/ddppo
cd experiments/ddppo
EXPT_DIR=$PWD
wget https://dl.fbaipublicfiles.com/habitat/data/baselines/v1/ddppo_objectnav_habitat2022_challenge_baseline_v1.pth

• Run baseline evaluation

cd <habitat-lab-path>

python habitat_baselines/run.py \
    --exp-config habitat_baselines/config/objectnav/ddppo_objectnav_hm3d.yaml \
    --run-type eval \
    EVAL.SPLIT "val_mini" \
    EVAL_CKPT_PATH_DIR $EXPT_DIR/ddppo_objectnav_habitat2022_challenge_baseline_v1.pth \
    NUM_ENVIRONMENTS 2 \
    TENSORBOARD_DIR $EXPT_DIR/eval_minival \
    LOG_FILE $EXPT_DIR/log_eval_minival.txt \
    VIDEO_OPTION "['disk']" \
    VIDEO_DIR $EXPT_DIR/video_minival

[arjung128]

@xuai05 I’m getting this when I install habitat-sim with conda install habitat-sim-challenge-2022 -c conda-forge -c aihabitat. It seems to go away when I install habitat-sim with conda install habitat-sim withbullet headless -c conda-forge -c aihabitat. But with this workaround, the semantic annotations don’t seem to get downloaded…