October #1
In October, I mainly working on researching into existing technologies that have implications for sign language movement detection (through camera).
Image Recognition:
- Basically mature nowadays. My mentor has the technology himself.
- Photos taken by normal camera (e.g. on smart phones) can be used
- Able to process thousands of data/images now
- After collecting data, the processing is easy to be operated on common computational models
- can add fish-eye lens (universal 3-in-1 clip-on camera lens) --> wider angle
Video Recognition:
- currently allows only tens of videos
- requires depth camera to collect data (e.g. Kinect)
- requires professional computers to process data **(depth analysis? need more research)
Human Pose Estimation
- requires depth camera to collect data (e.g. Kinect)
- requires professional computers to process data
- need more research
-
(Articulated) Hand Tracking
- requires depth camera to collect data (e.g. Kinect)
- requires professional computers to process data
- need more research
- e.g. https://youtu.be/A-xXrMpOHyc
|
Handpose: Articulated Hand Tracking |
|
Kinect 3D Hand Tracking |