Datasets
Human Action Multi-Modal Monitoring in Manufacturing (HA4M) Dataset
The HA4M dataset is a collection of multi-modal data of actions by different subjects in two assembly scenarios for manufacturing. It provides a good test-bed for developing, validating and testing techniques and methodologies for the recognition of assembly actions.
Six types of simultaneous data are supplied: RGB frames, Depth maps, IR frames, RGB-Depth-Aligned frames, Point Clouds and Skeleton data.
Data from vines before and after trimming (June 2022)
The dataset is designed to test automatic algorithms for measuring changes in the canopy of grapevine plants using natural, on-field images.
A Microsoft Azure Kinect DK sensor monitored two segments of vine rows before and after a trimming practice to produce 36 captures including color, depth, infrared images, and point clouds.
The dataset also includes manual annotation.
E-crops @ San Donaci - Vineyard Dataset (2021)
This dataset is made of natural images acquired from a moving vehicle by the Intel RealSense D435 RGB-D camera, in a commercial field in San Donaci (Apulia, Italy). The grape variety was Vitis vinifera, cultivar “Negroamaro” (red grape variety).
The dataset is made of 315 multimodal images (RGB and aligned depth), manually labeled for semantic segmentation.
Labeled images from vineyards in Switzerland
This dataset is made of natural images acquired in-field by the Intel RealSense R200 (Santa Clara, CA, USA) RGB-D camera. The resolution of the color images has been set to 640×480 pixels as the maximum achievable resolution of the depth maps (not included in this dataset).
Images are segmented according to 5 classes: bunch, pole, wood, leaves and background.
Short Physical Performance Battery Dataset
This dataset includes data (sex, age, skeletal joint positions) of patients (both healthy and affected by neurodegenerative diseases) performing the Short Physical Performance Battery (SPPB) tests.
Subjects were grabbed by three low-cost surveillance cameras. Then, proper video processing techniques computed the skeletal joints of the subjects.
The dataset is released as Matlab .mat archives.