Eye tracking work

He continued to work on computer modeling of eye traking this week : he optimized the algorithm, now he handles the 1350x768 image on the vaio tz laptop not in 8 minutes, as before, but in 2. There are a couple more ideas on how to speed up the work. Purpose: image processing seconds in 30.

Well, the video fixation points after the sacadas:


')
This picture was taken:


^{"Velobike: original"}

And the result of processing, provided that the distance to the monitor is ~ 40 centimeters, the monitor is 20 '' with a resolution of 1680x1050


^{"Velobike: processing result"}

The video above shows how the algorithm fixes the points of gaze fixation. The duration of fixation (reflected on the heatmap) is calculated as the density of lines over the area.

PS
For the sake of curiosity, how do other people model the distribution of attention? Here is a video of the simulation found in YouTube:



It can be seen that models something differently :) Judging by the comments on this and other videos , a specially trained neural network is used. But my algorithm has the ability to predict sakad :)