a Shows the camera image ( red stack, green stack, and blue wire roll). The objects and the linguistic input are the same as that of Demonstration 2a but the blue target object location is shifted upwards in the image. The robot answers “Where is the blue object relative to the green one?” by selecting the spatial term “above”. Smaller arrows indicate activity flow in the direction opposite to that of the dominant flow of the taskĭemonstration 2b neural field activity just before response generation. a Shows the color-term nodes activity, with the red color-term node ( red line) triggered by the red color-space field activation. The localized activation is elevated, leading to a detection instability ( ellipse). b Shows the activation time course of the red color-space field (projected onto the horizontal axis). Activity is elevated in the region to the right of the reference object ( upper region), leading to the detection instability ( ellipse) and activation into the color-space fields ( upward arrow). c Shows the right spatial semantic field time course (projected onto the horizontal axis as in Fig. 5). The right node ( red line) passes through a detection instability ( ellipse), boosting the right semantic field ( upward arrows). The right input indicates the time point of linguistic input at the start of the trial. d Shows the spatial-term node activation over the trial ( horizontal axis, seconds vertical axis, activation). x sp is the horizontal axis of the image planeĭemonstration 1b time course. Here,1 s ≈4 ♱0 3 integration time− steps. We measured time in seconds to maintain consistency across all plots in the present work. Smaller arrows indicate activity flow in the direction opposite to that of the dominant flow of the task. The right spatial semantic field activity boosts the activity of the right node ( red line), pushing it through the detection instability ( ellipse), triggering the response. d Depicts the activation profile for the spatial-term nodes. Once activity surpasses the detection instability ( ellipse) it propagates activation to the linked right spatial-term node. Color-space field activity leads to a localized activation profile for the blue object location ( middle portion of field). c Depicts the time course of the right spatial semantic field with activity projected onto the horizontal axis in the same manner described for ( b). When activity in the blue color-space field reaches the detection instability ( ellipse), that field passes activation into the spatial semantic fields ( downward arrow). Along the vertical axis of b, the lower portion corresponds to the leftmost image region, upper portion the rightmost image region. b Depicts the time course of the projection of the blue color-space field activity onto the horizontal axis over the trial. This increases activity of the blue color-term node ( red line), causing a detection instability ( ellipse) and activity propagation from the blue node to the blue color-space field ( downward arrow). The blue input indicates the time point of linguistic input into the node. a Shows the color-term nodes activity over the trial (the horizontal axis represents time, the vertical axis represents activation). This increases activity of the linked node, triggering the robot’s answer rightĭemonstration 1a time course. This leads to a localized positive activation in the right field ( red arrow) at the location of the blue target. The active regions in the color-space fields propagate activity to the spatial semantic fields. d Shows the spatial semantic field activation profiles after the shift of semantic templates to the reference object location. The blue color-term node input specifying the target object uniformly raises the activation of the entire blue color-space field, leading to an activation peak at the blue object location. c Depicts the color-space field activations induced by the current scene. b Shows the reference field activation corresponding to the green reference object selected by user. a Shows the camera image ( green toothpaste tube, a blue wire roll, and a red plastic apple). The robot answers the question “Where is the blue object relative to the green one?” by selecting right. Neural fields activity just before response generation.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |