Skip to main content

Advertisement

Table 7 Effect of changing the number of training samples

From: First-person reading activity recognition by deep learning with synthetically generated images

  Number of training samples for each class    
  1000 5000 25,000  
  Others Reading Others Reading Others Reading  
Scores Precision 73.3 89.2 70.8 92.9 70.1 96.2
for each class Recall 46.0 98.4 69.9 95.0 84.2 92.0
  F-measure 52.9 93.3 64.9 93.5 72.0 93.4
Averages Precision 81.2 81.9 83.1
over classes Recall 72.2 82.5 88.1
  F-measure 73.1 79.2 82.7
  1. Best methods for a given measure are specified in italic type