Parameter Adjustment in the Supermarket Scenario

In the supermarket scenario, the Supermarket_Seg_RGBSuction picking configuration folder should be used.

It is recommended to use a GeForce GTX 10 Series graphics card with a memory of at least or above 4G when you use the model for the supermarket scenario. When you run this Step for the first time, the deep learning model will be optimized according to the hardware type and the one-time optimization process takes about 15 to 35 minutes.

Parameter Adjustment Order

  1. Set a Working Distance first to limit the range for the camera to capture the data.

  2. Adjust the Contour Detection (Smart Placement, Special Shape) parameter to determine whether to process objects whose contours are detected with priority.

  3. The objects in the supermarket scenario are usually stacked randomly, adjusting the Overlap Detection parameter group can divide the objects into objects that are overlapped and not overlapped.

  4. Adjust the Pose Processing parameter group to control the upward offset along the Z-axis to avoid damaging the object.

  5. Adjust the Suction Cup Configuration parameter group to divide the suction cups into sections.

  6. Adjust the Pose Sorting Logic parameter group to determine the picking sequence.

Parameter Description

Working Distance

Minimum Working Distance
Description: This parameter is used to specify the minimum distance (in millimeters) between the camera and the object in the scene. If a bin is used, this parameter should be the distance from the camera to the top edge of the bin. Please set the value according to the actual situation.
Default value: 0 mm
Value range: 0–3000 mm
Maximum Working Distance
Description: This parameter is used to specify the maximum distance (in millimeters) between the camera and the object in the scene. If a bin is used, this parameter should be the distance from the camera to the bottom edge of the bin. Please set the value according to the actual situation.
Default value: 3000 mm
Value range: 0–3000 mm

Contour Detection (Smart Placement, Special Shape)

Prioritize Contours:
Description: If this feature is enabled, the objects whose contours are detected will be processed first in the further calculation.
Default value: Selected
Tuning recommendation: If the project does not have a high requirement for the overlap detection, you can enable this feature to increase the picking success rate and avoid changing bins frequently. If there is a high requirement on the overlap detection and only the objects that are NOT overlapped should be picked, it is recommended to disable this feature.

Overlap Detection

Max Number of Objects
Description: This parameter is used to specify the maximum number of objects that will be considered overlapped.
Default value: 6
Value range: 0–10
Tuning recommendation: Please set the value according to the actual situation.

Pose Processing

Upward Offset along Z-Axis
Description: This parameter is used to specify the offset distance (in millimeters) along Z-axis above the object to prevent the robot from squeezing the object.
Default value: 0
Value range: 0–10 mm
Tuning recommendation: Please set the parameter according to the actual situation.
Max Number of Output Poses
Description: This parameter is used to specify the maximum number of output poses.
Value range: 0–20

Suction Cup Configuration

Allocate by Mask Heftiness
Description: This parameter is used to allocate objects to different groups according to the heftiness of their masks, and the suction cup configurations for each group are different.
Value list:
  • Disable (Default value): Do not allocate objects to multiple groups according to the mask heftiness.

  • Into Two Groups: Allocate the objects to two groups according to Heftiness Threshold 1.

  • Into Three Groups: Allocate the objects to three groups according to Heftiness Threshold 1 and Heftiness Threshold 2.

Heftiness Threshold 1 /Heftiness Threshold 2:
Description: This parameter is used to set the heftiness threshold (in millimeters) for sorting masks.
Default value: 0mm
Value range: 0–200mm

Note

  • When Allocate by Mask Heftiness is set to Into Two Groups, you will need to set Heftiness Threshold 1. The allocation logic is as follows:

    • If the mask heftiness of the object is less than Heftiness Threshold 1, the object will be put into the first group.

    • If the mask heftiness of the object is greater than Heftiness Threshold 1, the object will be put into the second group.

  • When Allocate by Mask Heftiness is set to Into Three Groups, you will need to set Heftiness Threshold 1 and Heftiness Threshold 2. The Heftiness Threshold 2 must be greater than Heftiness Threshold 1. The allocation logic is as follows:

    • If the mask heftiness of the object is less than Heftiness Threshold 1, the object will be put into the first group.

    • If the mask heftiness of the object is between Heftiness Threshold 1 and Heftiness Threshold 2, the object will be put into the second group.

    • If the mask heftiness of the object is greater than Heftiness Threshold 2, the object will be put into the third group.

Allocate by Mask Span
Description: This parameter is used to allocate objects to different groups according to the span of their masks, and the suction cup configurations for each group are different.
Value list:
  • Disable (Default value): Do not allocate objects to multiple groups according to the mask span.

  • Into Two Groups: Allocate the objects to two groups according to Mask Span Threshold.

Mask Span Threshold
Description: This parameter is used to set the span threshold (in millimeters) for sorting masks.
Default value: 0 mm
Value range: 0–200 mm

Note

When the Allocate by Mask Span is set to Into Two Groups, you will need to set this parameter. The allocation logic is as follows:

  • If the mask span of the object is less than Mask Span Threshold 1, the object will be put into the first group.

  • If the mask span of the object is greater than Mask Span Threshold 1, the object will be put into the second group.

Pose Sorting Logic

The detected poses in this Step will be sorted according to the pose score in descending order.

Pose Score = Object Height × Pose Height Weight + Suction Cup Size × Suction Cup Size Weight + Object Length × Object Length Weight

Pose Height Weight
Description: This parameter is used to set the weight of the pose height when the score of a pose is calculated. If this weight is set to a larger value, objects that are at a high height will be more likely to be picked first.
Default value: 3
Value range: 0–5
Suction Cup Size Weight
Description: This parameter is used to set the weight of the suction cup size when the score of a pose is calculated. If this weight is set to a larger value, objects that should be picked with larger suction cups will be more likely to be picked first.
Default value: 1
Value range: 0–5
Object Length Weight
Description: This parameter is used to set the weight of the object length when the score of a pose is calculated. If this weight is set to a larger value, objects with longer lengths will be more likely to be picked first.
Default value: 1
Value range: 0–5

Visualization

Note

The Visualization parameter group is only used for selecting what will be displayed in the Debug Output window and will not affect the adjustment of the above parameters.

Enable
Description: Once this option is selected, you can select the items you want to visualize in the Debug Output window.
Default value: Selected
Visualization Type
Description: This parameter is used to specify the items you want to visualize.
Value list:
  • Final Score

  • Show Pose Height

  • Show Suction Cup Diameter

  • Show Object Length

Default value: Final Score

Hint

This parameter will only be displayed after the visualization is enabled.