Recognizing Objects in Smart Homes Based on Human Interaction - Advanced Concepts for Intelligent Vision Systems - page 139

(a) (b) (c) (d)

(e) (f) (g) (h)

Fig. 4. Example images of some activities. (a) cutting; (b) scrambling; (c) eating; (d)
using microwave; (e) reading; (f) typing on computer; (g) sitting and watching TV; (h)
lying on sofa.

Fig. 5. The room layout and estimated grid object types shown in color: floor (gray),
chair (blue), sofa (green), worktop (cyan), microwave (magenta), TV (red), Dtable
(yellow), computer (orange)

Table 2. Precision and recall for AIR lab object recognition, in terms of the number
of recognized grids. Note that recall ∗ is not the recall of the algorithm, since the user's
activity does not cover all possible locations of the objects. But the recall number is
nevertheless calculated by dividing the number of recognized grids with the total num-
ber of grids occupied by the object. So recall ∗ is expectedly lower than the algorithm's
recall.

floor chair sofa worktop microwave Dtable TV computer
precision 0.88 0.64 0.8

1 1 0.78 0.27 0.5

recall ∗ 0.47 0.82 0.5 0.67 0.5 0.67 0.75 0.17