Keras implementation of the proposed single-image adversarial depth estimation model.
Example 1[1]:
Example 2[1]:
[1](Input, U-Net, DenseNet)
The model is written and tested on Ubuntu 16.04 using Keras 2.2.0 and Tensorflow 1.8.0. Matplotlib 2.2.2 is required for visualizing the various outputs.
The model was trained on the depth prediction split of the KITTI dataset (http://www.cvlibs.net/datasets/kitti/eval_depth.php?benchmark=depth_prediction). You can download the pre-trained models for each generator network in the links below:
- Autoencoder generator: https://drive.google.com/open?id=1pPcIENf_66RKZCPZ9xHEo3sqZc3VKBoi
- DenseNet generator: https://drive.google.com/open?id=1HAFEj3AVm6a4ZF-xAk9pBZnLXQoVcSH3
In-depth presentation of the proposed models at: https://www.dropbox.com/s/bmeg1yextjawl4b/Single_Image_Depth_Estimation_Using_Generative_Adversarial_Networks.pdf?dl=0 (Greek)