UnDEMoN

UnDEMoN: Unsupervised Depth and EgoMotion Network Madhu Babu V, Kaushik Das, Anima Majumder and Swagat Kumar in IROS 2018 (oral)

Prerequisites

This codebase was developed and tested with Tensorflow 1.0, CUDA 8.0 and Ubuntu 16.04.

Preparing training data

We have defined an order that the data has to be preprocessed before training.

For KITTI, first download the dataset using this script provided on the official website, and then run the prepare_train_data_stereo.py with the necessary arguments.

Training

Once the data are formatted following the above instructions, you should be able to train the model by running the train.py.

You can then start a tensorboard session by

tensorboard --logdir=/path/to/tensorflow/log/files --port=8888

and visualize the training progress by opening https://localhost:8888 on your browser. If everything is set up properly, you should start seeing reasonable depth prediction.

Testing

Download the checkpoints from here and use the test_kitti_depth.py and test_kitt_pose.py with necessary arguments.

Evaluation

We have used the depth evaluation scripts from monodepth and the pose evalution scripts from the SfMLearner. We are very thankful the Monodepth and SfMLearner authors for their code bases.

Reference

If you find our work useful in your research please consider citing our paper:

@inproceedings{babu2018undemon,
  title={UnDEMoN: Unsupervised Deep Network for Depth and Ego-Motion Estimation},
  author={Babu, V Madhu and Das, Kaushik and Majumdar, Anima and Kumar, Swagat},
  booktitle={2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  pages={1082--1088},
  year={2018},
  organization={IEEE}
}

👩‍⚖️ License

Non-Commercial use only.

olibd / undemon