This code provides a combination of DSO and Monodepth. For every keyframe, depth values are initialized with the prediction from Monodepth.

Absolute keyframe trajectory RMSE (in meter) on KITTI dataset (DSO and ORB-SLAM numbers are from CNN-SVO paper)

• Setup dependencies of DSO (https://github.com/JakobEngel/dso)

• Build TensorFlow C++ API (https://github.com/yan99033/monodepth-cpp/tree/master/Tensorflow_build_instructions). This is the hardest part!
• Build monodepth-cpp (https://github.com/yan99033/monodepth-cpp).
• Prepare Monodepth pre-trained model. You can freaze .ckpt or download the model trained on cityscapes and fine-tuned on kitti here.

• Download the repository.

    git clone https://github.com/muskie82/CNN-DSO.git
  

• Modify paths to include directories and libraries of TensorFlow and monodepth-cpp in CMakeLists.txt (4 lines of /abosolute/path/to/XXXXX).

• Build

    cd CNN-DSO
    mkdir build
    cd build
    cmake ..
    make -j4
  

In addition to original DSO command line, you should specify the path to pre-trained model by cnn.

	bin/dso_dataset \
		files=XXXXX/sequence_XX/image_0 \
		calib=XXXXX/sequence_XX/camera.txt \
		cnn=XXXXX/model_city2kitti.pb \
		preset=0 \
		mode=1

• Direct Sparse Odometry, Engel, Jakob, Vladlen Koltun, and Daniel Cremers, IEEE transactions on pattern analysis and machine intelligence 40.3 (2018): 611-625. (https://github.com/JakobEngel/dso)
• Unsupervised monocular depth estimation with left-right consistency, *Godard, Clément, Oisin Mac Aodha, and Gabriel J. Brostow. *, CVPR. Vol. 2. No. 6. 2017. (https://github.com/mrharicot/monodepth)
• CNN-SVO: Improving the Mapping in Semi-Direct Visual Odometry Using Single-Image Depth Prediction., Loo, S. Y., Amiri, A. J., Mashohor, S., Tang, S. H., and Zhang, H, arXiv preprint arXiv:1810.01011 (2018). (https://github.com/yan99033/CNN-SVO)
• stereo_dso by HorizonAD: https://github.com/HorizonAD/stereo_dso

GPLv3 license. I don't take any credit from DSO, Monodepth and monodepth-cpp. Please check them.