NeoNav

This is the implementation of our AAAI 2020 paper NeoNav: Improving the Generalization of Visual Navigation via Generating Next Expected Observations, training and evaluation on Active Vision Dataset (depth only).

Navigation Model

Implementation

Training

• The environment: Cuda 10.0, Python 3.6.4, PyTorch 1.0.1
• Please download "depth_imgs.npy" file from the AVD_Minimal and put the file in the train folder.
• Please download our training data HERE.
• Our trained model can be downloaded from HERE. If you plan to train your own navigation model from scratch, some suggestions are provided:
  • Pre-train the model by using "python3 ttrain.py" and terminate the training when the action prediction accuracy approaches 70%.
  • Use "python3 train.py" to train the NeoNav model.

Testing

• To evaluate our model, please run "python3 ./test/evaluate.py" or "python3 ./test/evaluete_with_stop.py"

Results

Start	End	Start	End

Contact

To ask questions or report issues please open an issue on the issues tracker.

Citation

If you use NeoNav in your research, please cite the paper:

@article{wuneonav,
  title={NeoNav: Improving the Generalization of Visual Navigation via Generating Next Expected Observations},
  author={Wu, Qiaoyun and Manocha, Dinesh and Wang, Jun and Xu, Kai}
}