This repository contains python code to evaluate on the Stereo-Endoscopic Reconstruction Validation dataset based on CT (SERV-CT).
SERV-CT contains 16 stereo-endoscopic image pairs with reference anatomical segmentation derived from CT. Two different ex vivo porcine samples were imaged using the straight and 30° endoscopes from the original classic da Vinci™ Surgical System (Intuitive Inc., Sunnyvale, US).
The dataset provides reference in both disparity and depth (Z distance between 3D point and camera center) domain as well as occlusion images.
To download the SERV-CT dataset, visit the the project's homepage.
The arXiv version of our paper can be found SERV-CT: A disparity dataset from CT for validation of endoscopic 3D reconstruction
We provide 16 stereo rectified image pairs with reference disparity and depth maps. For the second experiment (samples 9-16), in addition to CT reference samples, we provide reference samples from a structured light device and we call those samples RGB. The dataset is organised as as follows:
└── SERV-CT
├── Experiment_1 - frames 001-008
│ ├── Ground_truth_CT - The reference data using an O-arm CT scan
│ │ ├── DepthL - Left image depth maps (mm depth for each pixel scaled by 256 and stored as 16 bit PNG)
│ │ ├── DepthR - Right image depth maps (mm depth for each pixel scaled by 256 and stored as 16 bit PNG)
│ │ ├── Disparity - left-to-right disparity (pixel disparity scaled by 256 for subpixel accuracy and stored as 16 bit PNG)
│ │ ├── OcclusionL - Colour coded occlusion images
│ │ └── OcclusionR (Yellow - non overlap, blue - outside the reference surface, red - not visible in the right image)
│ ├── Left_rectified - left rectified images (720x576 24-bit colour PNG)
│ ├── Rectified_calibration - JSON calibration files containing P1, P2 and Q, (units are in pixels and mm)
│ └── Right_rectified - right rectified images (720x576 24-bit colour PNG)
└── Experiment_2 - frames 009-016
├── Ground_truth_CT
│ ├── DepthL
│ ├── DepthR
│ ├── Disparity
│ ├── OcclusionL
│ └── OcclusionR
├── Ground_truth_RGB - The reference data using a Creaform RGB scan
│ ├── DepthL
│ ├── DepthR
│ ├── Disparity
│ ├── OcclusionL
│ └── OcclusionR
├── Left_rectified
├── Rectified_calibration
└── Right_rectified
Because we currently do not collect email addresses when we provide a copy of the SERV-CT dataset, we will keep a changelog here. If you wish to get updates about the dataset or the evaluation code(changes/additions/fixes), press the watch button for this repository. Any changes made will be logged here.
- 1/2/21 Release of Evaluation code for the SERV-CT dataset. Code includes evaluation scripts for both disparities and depthmaps. Additionally we provide code to read and write disparity and depthmap files in the format used by the dataset.
- 18/1/21 Replaced an outdated version of the SERV-CT dataset. Download the latest from the project's webpage
- 22/12/20 The SERV-CT paper preprint is now available to download.
We used anaconda to manage dependencies for this project. To setup an anaconda environment suitable to running this code, after installing anaconda, create a new environment.
- cd to this project's directory
- create an anaconda enviroment using the requirements.txt file
conda create -n servcttk --file requirements.txt -c conda-forge
- activate the environement
conda activate servcttk
The toolkit provide two scripts for evaluation, one for depthmaps and one for disparities. The scripts assume that the the samples to evaluate are named after the samples they correspond to. For each domain [disparity, depthmap] a single folder containing all samples for evaluation should be created following the file structure bellow.
.
├── Disparities
│ ├── 001.png
│ ├── 002.png
│ :
│ ├── 015.png
│ └── 016.png
└── Depthmaps
├── 001.png
├── 002.png
:
├── 015.png
└── 016.png
Disparity is measured as the horizontal displacement of corresponding pixels in two views.
python -m scripts.evaluate_disparity /path/to/dataset/root/directory \
/path/to/directory/containing/disparity/samples/for/evaluation -v
Depth error is measured in mm distance along the Z camera axis.
python -m scripts.evaluate_depthmap /path/to/dataset/root/directory \
/path/to/directory/containing/depthmap/samples/for/evaluation -v
Both scripts will create a folder named error, under the directory evaluation samples
are stored, and export results for each dataset/modality/occlusion in csv format
in many different metrics. Furthermore, a separate .csv will be generated
summarizing the performance across every experiment using only a subset of metrics.
Depth error is measured in mm distance and disparity in pixel displacement between views in the horizontal dimension.
Please cite the following publication whenever research making use of the SERV-CT dataset or this repository's code is reported in any academic publication or research report:
@article{edwards2020serv,
title={SERV-CT: A disparity dataset from CT for validation of endoscopic 3D reconstruction},
author={Edwards, PJ and Psychogyios, Dimitris and Speidel, Stefanie and Maier-Hein, Lena and Stoyanov, Danail},
journal={arXiv preprint arXiv:2012.11779},
year={2020}
}