A method for modularizing high dimensional neural networks into neural networks of lower input dimensions. The method is suited to generating full-DOF robot grasping actions based on images of parts to be picked. In one example, a first network encodes grasp positional dimensions and a second network encodes rotational dimensions. The first network is trained to predict a position at which a grasp quality is maximized for any value of the grasp rotations. The second network is trained to identify the maximum grasp quality while searching only at the position from the first network. Thus, the two networks collectively identify an optimal grasp, while each network's searching space is reduced. Many grasp positions and rotations can be evaluated in a search quantity of the sum of the evaluated positions and rotations, rather than the product. Dimensions may be separated in any suitable fashion, including three neural networks in some applications.

A card reader includes a card lock mechanism for preventing pulling-out of a card from an insertion port when the card is jammed in a conveyance passage. The card lock mechanism includes a motor, a lock member driven by the motor to be moved between a contact position and a retreated position, a knob mechanically connected with the motor, and a detection mechanism for detecting the lock member located at the retreated position. The card reader includes a control part which is configured so that, after the motor is driven to move the lock member to the contact position, when the detection mechanism detects that the lock member has moved to the retreated position, the control part executes a tactile stimulation sequence in which the motor is driven to apply movement to the knob so that attention of an operator operating the knob is called in a tactile manner.

A method for modularizing high dimensional neural networks into neural networks of lower input dimensions. The method is suited to generating full-DOF robot grasping actions based on images of parts to be picked. In one example, a first network encodes grasp positional dimensions and a second network encodes rotational dimensions. The first network is trained to predict a position at which a grasp quality is maximized for any value of the grasp rotations. The second network is trained to identify the maximum grasp quality while searching only at the position from the first network. Thus, the two networks collectively identify an optimal grasp, while each network's searching space is reduced. Many grasp positions and rotations can be evaluated in a search quantity of the sum of the evaluated positions and rotations, rather than the product. Dimensions may be separated in any suitable fashion, including three neural networks in some applications.

A card reader includes a card lock mechanism for preventing pulling-out of a card from an insertion port when the card is jammed in a conveyance passage. The card lock mechanism includes a motor, a lock member driven by the motor to be moved between a contact position and a retreated position, a knob mechanically connected with the motor, and a detection mechanism for detecting the lock member located at the retreated position. The card reader includes a control part which is configured so that, after the motor is driven to move the lock member to the contact position, when the detection mechanism detects that the lock member has moved to the retreated position, the control part executes a tactile stimulation sequence in which the motor is driven to apply movement to the knob so that attention of an operator operating the knob is called in a tactile manner.

System and techniques for an image filter based on row identification are described herein. A crop row center represented in an image of a crop row can be calculated. A filter corresponding to a set of expected crop characteristics of the crop row can be obtained. Elements in the image can then be categorized based on applying the filter to the image when the filter anchored on the crop row.

A card reader may include a card insertion port; a card conveying passage; and a card lock mechanism structured to prevent drawing of the card from the card insertion port. The card lock mechanism may include a motor; a lock member comprising a prevention pawl structured to move between a contact position and a retreated position; and a power transmission mechanism structured to transmit power of the motor to the lock member. The power transmission mechanism may include a final gear which is disposed on a most lock member side in a transmitting direction of power from the motor to the lock member. The lock member may include a sector gear which is engaged with the final gear.

System and techniques for an image filter based on row identification are described herein. A crop row center represented in an image of a crop row can be calculated. A filter corresponding to a set of expected crop characteristics of the crop row can be obtained. Elements in the image can then be categorized based on applying the filter to the image when the filter anchored on the crop row.

System and techniques for an image filter based on row identification are described herein. A crop row center represented in an image of a crop row can be calculated. A filter corresponding to a set of expected crop characteristics of the crop row can be obtained. Elements in the image can then be categorized based on applying the filter to the image when the filter anchored on the crop row.