This is a dataset we generated for our IPCAI 2019 submission "Implicit Domain Adaptation with Conditional Generative Adversarial Networks for Depth Prediction in Endoscopy".
The dataset consists of 16,016 RGB images with corresponding ground truth depth. The images were resized to 256 x 256 pixels. The depth is scaled to [0,1] which corresponds to [0,20] cm. The data is divided into groups according to its texture (T1, T2, T3) and the lighting (L1, L2, L3). For each configuration there are four to five different subsets generated by randomly shifting and rotating the virtual camera.
Click here to view and download the dataset. (Password no longer required)
If you found our dataset helpful for your research please consider citing us:
@article{rau2019implicit,
title={Implicit domain adaptation with conditional generative adversarial networks for depth prediction in endoscopy},
author={Rau, Anita and Edwards, PJ Eddie and Ahmad, Omer F and Riordan, Paul and Janatka, Mirek and Lovat, Laurence B and Stoyanov, Danail},
journal={International journal of computer assisted radiology and surgery},
pages={1--10},
publisher={Springer}
}
Find our paper here, return to our homepage, or see what the rest of our team is working on.
