Instruction-Aligned Hierarchical Waypoint Planner for Vision-and-Language Navigation in Continuous Environments

Preparation

1. Install habitat-lab and VLN-CE

2. Install other requirements:

   pip install -r requirements.txt

3. Prepare datasets

   • R2R_VLNCE_v1-2
     • R2R_VLNCE_v1-2.zip
     • R2R_VLNCE_v1-2_preprocessed.zip
   • ddppo

   +-data
   |   +-checkpoints
   |   +-datasets        
   |   |   +-R2R_VLNCE_v1-2_preprocessed
   |   +- ddppo-models    
   

Train, evaluate and test

1. train_IAW.bash is the script for training
2. eval_IAW.bash is the script for evaluating
3. inference_IAW.bash is the script for inference
4. The weight with the best performance on VLN-CE val unseen can be downloaded here (Access code: iahwp).

Citation