1. Feature Introduction

In Deep Learning models, Shadow Mode is a data enhancement method that reuses real data accumulated during model operation in actual scenes for model training, increasing data diversity and improving the model's generalization capability.

2. Applicable Scenarios

When the on-site system has already achieved a detection success rate of more than 95% for the same type of object and the goal is to reach 99.9%, you can train the model yourself using Shadow Mode.

3. Operating Instructions

1. Make sure the total amount of shadow data meets the requirement, with at least 100 shadow data samples

2. On the main interface, click workpiece, click a specific workpiece to enter the workpiece configuration interface, and scroll down to Shadow Mode

3. Task Environment

   Function Description

   Click Edit Task Environment to view and edit shadow data samples without workpiece instance information. In the task environment data, you need to delete sample data that does not contain any workpieces and use it as the task environment (in very rare cases, samples may contain unrecognized workpieces; such samples will mislead the model and need to be deleted, see the lower right image for reference), ensuring that the task environment data contains no workpieces and only bottom supports/pallets/ground, for subsequent shadow training and to prevent the model from mistakenly recognizing non-workpiece objects such as bottom supports/pallets/ground.

The sample requirement is a completely undetectable 2D recognition error sample. 2D error samples or 3D error samples filtered by functions will not be displayed or take effect in task environment editing.

Applicable Scenarios

There is a risk of missed detection in the model, and it is necessary to add task environment data of bottom supports/pallets/ground to enhance training and help the model clarify the recognized objects.

4. Select the model to be trained

5. Data storage location

   • After instance filtering: selected by default.

     Applicable Scenarios: Using the data output from "After instance filtering" for training is suitable for cases where incorrect segmentation exists and filtering is required.

   • Before instance filtering

     Applicable Scenarios: Using the data output from "Before instance filtering" for training is suitable for scenes where the workpiece segmentation effect is good

6. Keep the number of iterations at the default value

7. Set the model name

8. Set remarks, which can be left blank

   

   

9. Click Start Training

10. After training ends, you can directly select the model for use under Vision Model.

4. Frequently Asked Questions

1. Storage path of shadow data

   • Storage path after instance filtering: \项目文件夹\data\ShadowDate\product_{工件id}\InstanceCriteria

   • Storage path before instance filtering: \项目文件夹\data\ShadowDate\product_{工件id}\InstanceBuilder

   

   

2. How to confirm whether Shadow Mode is enabled

   

3. How much total shadow data is required before training can be enabled

   

4. The trained model performs poorly. How should this be handled?

   • Check whether the detected material and the material used during Shadow Mode training belong to the same category

   • In the workpiece interface, check whether the workpiece model has been replaced with the latest trained Deep Learning model

   • In the vision Parameter interface, try adjusting Scaling Ratio in 2D recognition

   • When a single scaling ratio cannot meet the actual scene requirements (for example, the optimal scaling ratio of objects on the top layer and bottom layer in a depalletizing scene may be different), select the Auto Enhance function in the vision Parameter interface and set multiple Auto Enhance-Scaling Ratios. For details, see Depalletizing Vision Parameter Adjustment Guide

Previous pageTrain Imaging Models via One-Click Linking

Next pageHow to Use KINGFISHER Shadow Mode