Mech-Vision 1.7.2 Release Notes¶
This document introduces the new features, improvements, and resolved issues of Mech-Vision 1.7.2.
New Features¶
Medicine Boxes Scenario Is Added to the “Predict Pick Points V2” Step¶
A new “Medicine Boxes” scenario is added to the “Predict Pick Points V2” Step, which enables to sort medicine boxes that are stacked randomly.
You can find the corresponding “Medicine Boxes” project with the deep learning model in the Solution Library.
New Feature in “Deep Learning Model Package Management Tool”¶
In the new “Deep Learning Model Package Management Tool”, the model package’s inference efficiency can be configured by adjusting the “Batch size” and “Precision” parameters. Only models exported by Mech-DLK 2.4.1 or above support this feature.
Hint
It is recommended to use deep learning model packages exported by Mech-DLK 2.4.1 or above with Mech-Vision 1.7.2 or above.
However, some deep learning model packages exported by Mech-DLK 2.4.1 can be used in Mech-Vision 1.7.1. Please pay attention to the compatibility issue.
New Step “Deep Learning Model Package Inference”¶
From Mech-Vision 1.7.2, “Deep Learning Model Package CPU Inference” and “Deep Learning Model Package Inference (Mech-DLK 2.2.0+)” are merged into one “Deep Learning Model Package Inference” Step.
If projects that are created with the software of previous versions are opened in Mech-Vision 1.7.2, the “Deep Learning Model Package CPU Inference” and “Deep Learning Model Package Inference (Mech-DLK 2.2.0+)” Steps will be automatically replaced with the “Deep Learning Model Package Inference” Step.
This Step performs inference with single model packages or cascaded model packages exported by Mech-DLK and outputs the inference result. This Step only supports model packages exported by Mech-DLK 2.2.0 or above.
Note
From Mech-DLK 2.4.1, model packages can be divided into single model packages and cascaded model packages.
Single model package: There is only one deep learning model in the model package, such as an “Instance Segmentation” model.
Cascaded model package: Multiple models are cascaded in the model package, and the output result of the previous model is input to the next model. For example, there are two models, “Object Detection” and “Instance Segmentation” in the model package and the inference sequence is . The output of “Object Detection” is input to the “Instance Segmentation” model.
When a cascaded model package is used for inference in this Step, the “Deep Learning Result Parser” Step can be used to parse the inference result of the cascaded model package.
New Step “Deep Learning Result Parser”¶
Mech-Vision 1.7.2 has introduced the “Deep Learning Result Parser” Step, which can parse the cascaded model package’s inference result exported by the “Deep Learning Model Package Inference” Step.
New Solution/Projects Added to the Solution Library¶
The following solution and projects have been added to the Solution Library in Mech-Vision 1.7.2.
Solution: Brake Discs (Single Station)
Project: Medicine Boxes, General Workpiece Recognition, Randomly Stacked Small Workpieces (Bolts)
Hint
If you need to use newly added built-in projects or solutions, please use the latest software.
Improvements¶
Algorithm Improvement¶
The algorithms of 3D matching Steps (“3D Coarse Matching”, “3D Fine Matching”, etc) in Mech-Vision 1.7.2 were improved, and the Step running speeds were enhanced.
Updated Step and Parameter Names¶
The following Step and parameter names have been updated in Mech-Vision 1.7.2.
Before Mech-Vision 1.7.2 |
Mech-Vision 1.7.2 |
|
Step Name |
Adjust 3D Poses by 2D Poses |
Convert 2D Poses to 3D Poses |
Parameter Name |
Filter Candidate Poses by Specified Axis Angles (in “3D Coarse Matching V2”, “3D Fine Matching”, etc) |
Filter Poses by Model Rotation Angle |