Evaluation Dataset (Depth Estimation)

NYU-Depth V2

#Metrics-RMSE, RMSE Log, δ1, δ2, δ3, Abs Rel, Sq Rel

Data description：

The NYU-Depth V2 data set is comprised of video sequences from a variety of indoor scenes as recorded by both the RGB and Depth cameras from the Microsoft Kinect. The collection locations of the data set are also very rich, including 464 scenes in 3 cities.

Dataset structure：

Amount of source data：

The dataset is divided into training set(35,599), validation set(654), test set(654).

Data detail：

Raw image: RGB image Depth map: Grayscale image

Sample of dataset：

Raw image：

Depth map:

Citation information：

@inproceedings{silberman2012indoor,
  title={Indoor segmentation and support inference from rgbd images},
  author={Silberman, Nathan and Hoiem, Derek and Kohli, Pushmeet and Fergus, Rob},
  booktitle={European Conference on Computer Vision},
  pages={746--760},
  year={2012},
  organization={Springer}
}

KITTI Eigen split

#Metrics-RMSE, RMSE Log, δ1, δ2, δ3, Abs Rel, Sq Rel

Data description：

The KITTI dataset consists of several outdoor scenes captured by on-board cameras and depth sensors of moving vehicles. This dataset utilizes the training/test set allocation scheme proposed by Eigen et al. The scheme covers 56 scenes, including 23,158 training image pairs and 652 test images.

Dataset structure：

Amount of source data：

The dataset is divided into training set(22,506), validation set(652), test set(652).

Data detail：

Raw image: RGB image Depth map: Grayscale image

Sample of dataset：

Raw image：

Depth map:

Citation information：

@article{eigen2014depth,
  title={Depth map prediction from a single image using a multi-scale deep network},
  author={Eigen, David and Puhrsch, Christian and Fergus, Rob},
  journal={Advances in neural information processing systems},
  volume={27},
  year={2014}
}

Licensing information：

TERMS OF USE