SELMA Dataset: SEmantic Large-scale Multimodal Acquisitions in Variable Weather, Daytime and Viewpoints

SELMA is a novel synthetic dataset for semantic segmentation in the autonomous driving context that contains more than 30K unique waypoints acquired from 24 different sensors including RGB, depth, semantic cameras and LiDARs, in 27 different weather and daytime conditions, for a total of more than 20M samples.

Kinect data acquisitions under different noise and light conditions

Test data used in the paper Joint Denoising and Interpolation of Depth Maps for MS Kinect Sensors. It includes several kinect RGB + depth acquisitions in different environments and light conditions. It was used to test the efficiency of denoising.

Multiple JPEG compression detection dataset

Multiple JPEG compression detection dataset. The dataset and the software were used to implement a multiple JPEG compression detector, which was presented in the paper "Discriminating multiple JPEG compressions using first digit features".

Time-Of-Flight and camera dataset

Data for super resolution applications: Three different scenes have been acquired with a camera and ToF setup. Calibration data is also available. The data has been used for the paper "Edge-preserving interpolation of depth data exploiting color information"

Time-Of-Flight and stereo dataset (dataset 1, with ground truth)

Time-Of-Flight and stereo dataset (dataset 2)

Time-Of-Flight and stereo dataset (dataset 3, with ground truth)

Time-Of-Flight and stereo dataset (dataset 4, with ground truth)

Data for ToF-stereo fusion applications: different scenes have been acquired with two cameras and ToF setup. Calibration and ground truth data are also available. Ground truth data is also available. The data has been used for the papers  "Locally Consistent ToF and Stereo Data Fusion", "A Probabilistic Approach to ToF and Stereo Data Fusion, and "Probabilistic ToF and Stereo Data Fusion Based on Mixed Pixels Measurement Models".

Color and depth dataset for joint color and depth segmentation.

Different scenes acquired with different sensors are available. The data has been used for the papers "Fusion of Geometry and Color Information for Scene Segmentation" and "Scene Segmentation from depth and color data driven by surface fitting" .

Multi-view dataset for multi-view plus depth image coding

This page contains the "kitchen" dataset used for the experimental results of the paper: "A Novel Multi-View Image Coding Scheme based on View-Warping and 3D-DCT"

In this page there are different datasets of hand gestures both for static and dynamic scenarios.

Markerless motion capture (external link)

Datasets for the paper  "Marker-less Motion Capture of Skinned Models in a Four Camera Set-up using Optical Flow and Silhouettes"