Dataset

Overview:

This dataset includes Gelsight videos collected using a Gelsight sensor on 93 out of 100 materials in the HaTT dataset as well as a few extra textures available at MPI's Haptic Intelligence Lab at the time of data collection.

We sincerely thank Prof Katherine Kuchenbecker and Dr. Yasemin Vardar for giving us access to the textures used in the Penn Haptic Texture Toolkit (HaTT) and hosting N. Heravi at the Max Planck Institute for Intelligent Systems (MPI-IS) during data collection.

Dataset can be downloaded here:

http://download.cs.stanford.edu/juno/HapticTexture/Sensor1(OldSensor).zip

Please reach out to nheravi@stanford.edu for any questions regarding the dataset.

Dataset Description:

The sensor used in this data collection is referred to as “Sensor 1” in the folders. Image obtained from Prof Yuan's webpage.

Sensor 1 Data:

Two sets of videos were collected using this sensor:

(1) Orientation Variation: The orientation of the sensor with respect to the material was varied while the sensor was being pressed approximately at the center of the material square. Three different variations in the orientation were conducted: (i) The sensor was approximately pressed parallel with the centerline of the material (ii) The sensor was approximately pressed parallel to one of the diagonals (ii) The sensor was approximately pressed parallel to the other diagonal. The naming convention of these videos files is: “Materialname_Orientation_Date_Time.avi” .

(2) Location Variation: The location of the sensor with respect to the material was varied while the sensor was being pressed approximately parallel with the centerline of the material. 5 different approximate locations (center and four corners) were chosen. . The naming convention of these videos files is: “Materialname_Date_Time.avi”. Data collected using this variation was used for obtaining the results in our ICRA 2020 paper as they were best fit for our specific application.