Loading...

Synthetic Video Generation Using Test Scene and Subject to Improve Fall Detection Accuracy

Moharamkhani, Armin | 2021

389 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 54430 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Amini, Arash; Mohammadzadeh, Nargesolhoda
  7. Abstract:
  8. Falling is a prevalent event among elderly people, which sometimes leads to their death. Automatic detection of fall can significantly reduce the resulting damages.Fallings can be detected using various modalities, among which we choose RGB videos captured by CCTV cameras because of its advantages. Due to the great advances in deep learning-based image/video classification methods, we focused on using these methods for fall detection. One of the main challenges in using deep learning methods is lack of enough training data. Unlike other activities, there are not enough falling samples available which is due to its unconscious nature. Moreover, simulating falling by actors can endanger their health. So in order to increase the accuracy of our fall detection system, we generated synthetic fall videos. We used deep neural networks and animation software for the purpose of synthesizing fall videos. We performed 2D pose and 3D joint location estimation from real videos using deep neural networks. We then used the resulting information to synthesize new fall videos under various viewpoints by means of animation softwares.We designed several experiments to show the benefit of using synthesized videos in training fall detection systems. We generated 24 videos from each existing real video in different viewpoints, which increased the classification accuracy by at least 50 % in the most of the experiments
  9. Keywords:
  10. Vision ; Neural Networks ; Human Pose Estimation ; Surveillance Cameras ; Deep Networks ; Dataset Gathering ; Free Viewpoint Video ; Three Dimentional Joint Location ; Eldery Fall

 Digital Object List

 Bookmark

No TOC