Data


Sample data from the FG2020 paper


This is a small data batch from the FG2020 paper "Towards a Visual Sign Language Dataset for Home Care Services". The batch includes a video recorded using a ximea camera, a video using a kinect2 rgb camera, the annotation files that were produced from elan, the json file that contains the keypoints as well as the script text and glosses in docx and xslx format respectively. The batch contain data from signer 8, signing the 5th script of our collection of Sophia Scripts.


File Description

  • n8_s5_t1_rgb.mp4 is the rgb video recorded with kinect 2 rgb sensor in mp4 format, 30 fps. Where n8 reffers to the 8th signer, s5 reffers to the 5th script from the collection of Sofia Scripts.
  • n8_s5_t1_depth.mj2 is the video recorded with kinect 2 depth sensor in mj2 format, and contains depth information. Where n8 reffers to the 8th signer, s5 reffers to the 5th script from the collection of Sofia Scripts.
  • n8_s5_t1.avi is the video recorded with the ximea camera in avi format, ~60 fps. Where n8 reffers to the 8th signer, s5 reffers to the 5th script from the collection of Sofia Scripts.
  • n8_s5_t1.eaf is the ELAN file that contains the annotation with timestamps. The annotation is based on n8_s5_t1.avi and the greek glosses from Sophia_Stress_script_5.xlsx. In order to use this file, you need to edit the line that reffers to the video path.
  • n8_s5_t1.json contains the hand keypoints that were extracted from the video.
  • n8_s5_t1_res.json contains the face keypoints that were extracted from the video.
  • Sophia_Stress_script_5.xlsx contains the greek glosses of the 5th script from the Sofia Scripts collection.
  • Sophia_Stress_script_5.docx contains the greek text manuscript of the 5th script from the Sofia Scripts collection.

Preview of the kinect 2 video


Sample videos of the recording in Athens

We present a small sample of the Greek Sign Language video dataset that was shot in Athens, from the 23rd to the 25th of July. The Greek glosses and the text of the script in Greek appear in the subtitles. The complete dataset will be publicly available after the video recordings and the annotation process are complete.