Before You Upgrade

1.6.2 Upgrade Notes

This section introduces the upgrade notes for Mech-Vision 1.6.2.

Mech-Vision Upgrade Notes

Notes for “Predict Pick Points” Steps

Mech-Vision 1.6.2 supports two “Predict Pick Points” Steps, which are “Predict Pick Points (Single Object Type)” and “Predict Pick Points (Any Objects)”. You do not need to install the Mech-Mind Software Environment to run these two Steps. However, you will need to use a deep learning server with a port number of 60000 or above.

If you are using Mech-Vision 1.5.3 or earlier versions, some deep learning Steps require that the Mech-Mind Software Environment or a special Mech-Vision version should be installed, and a deep learning server with a port number below 60000 should be used. Please refer to Compatibilities of Deep Learning Steps for detailed information.

In order to avoid environment conflict, it is prohibited to start a deep learning server with a port number below 60000 and another one with a port number of 60000 or above at the same time. Therefore, please pay attention to the following compatibility issues when you use these two Steps:

• When you create a new project and add the “Predict Pick Points” Step in the project, do not add other deep learning Steps that require the Mech-Mind Software Environment to be installed.

• If you have already opened a project containing the new “Predict Pick Points” Step in Mech-Vision, opening another project containing deep learning Steps that require the Mech-Mind Software Environment may cause environment conflict, and the projects cannot be run. To resolve the environment conflict, you can replace the deep learning Steps of earlier versions with the “Deep Learning Model Package Inference (Mech-DLK 2.2.0+)” or “Deep Learning Model Package CPU Inference” Steps. After replacing the Steps, you should train and import the deep learning model again.

1.6.1 Upgrade Notes

This section introduces the upgrade notes for Mech-Vision 1.6.1, Mech-Viz 1.6.1, and Mech-Center 1.6.1.

Mech-Vision Upgrade Notes

Notes for “Pick Anything” Projects

Currently, only Mech-Vision 1.6.1 special version supports the “Pick Anything” typical application projects. If you need to use “Pick Anything” projects, please contact Mech-Mind pre-sales engineers or sales to obtain the installation package of the special version.

When a “Pick Anything” project is run on the new version for the first time, the model will be converted. The higher the hardware configuration, the longer time the model conversion takes.

The new version is not compatible with “Pick Anything” projects created on an earlier version. If you want to use old “Pick Anything” projects on the new version, you need to delete and re-add the “Estimate Pick Points” Step.

V4 Cameras Should Be Used with Mech-Vision 1.6.0 or Later

Starting from 1.6.0, Mech-Vision is compatible with V4 cameras. Therefore, V4 cameras should be used together with Mech-Vision 1.6.0 or later.

Usage Notes for LSR V4/DEEP V4 Cameras

After upgrading Mech-Vision to 1.6.1, pay attention to the following notes when you are using LSR V4/DEEP V4 cameras:

• When opening old projects that use LSR V4/DEEP V4 cameras in Mech-Vision 1.6.1 for the first time, you need to reconnect the camera for Step “Capture Images from Camera” so as to make the “Camera Model” and “2D Image Type” parameters available.

• If the cameras used in existing projects have been upgraded to LSR V4/DEEP V4 cameras and these projects have used deep learning functions, you need to set the “2D Image Type” parameter to “Internal IR Image” for Step “Capture Images from Camera.” If you want to use “External Color Image” for image capturing instead, you need to retrain the models.

• If LSR V4/DEEP V4 cameras are used in new projects, especially if DEEP V4 cameras are used in depalletizing projects, you need to set the “2D Image Type” parameter to “External Color Image.” You can choose the 2D image type according to your project requirements in other situations.

Usage Notes for Step “3D Fine Matching”

• Mech-Vision 1.6.1 improved the standard deviation algorithm for Step “3D Fine Matching”, which makes the matching result filtering stricter. If you find that the detection rate decreases, you can manually set the “Standard Deviation” parameter to a lower value (0.003 m is recommended).

• It is not recommended to enable the Speed Up on Large Object Quantities option when only single object is in the scene. When this option is enabled, recognizing the single object will take longer.

Notes Related to Deep Learning

Before Mech-Vision 1.6.0, inference can be carried out only when a proper deep learning environment is installed. While Mech-Vision 1.6.0 and later have built-in inference models, inference can be carried out without installation of any deep learning environment.

On the industrial PCs or PCs without the deep learning environment, Mech-Vision 1.6.1 supports only the following deep learning Steps:

• Deep Learning Model Package Inference (DLK 2.2.0+): supports the models trained by Mech-DLK 2.2.0 or later

• Deep Learning Model Package CPU Inference: supports the models trained by Mech-DLK 2.2.1 or later

If an earlier project has already used deep learning Steps (other than above mentioned ones), you need to pay attention to the following notes to ensure that the earlier project can run normally after Mech-Vision is upgraded to 1.6.1:

• If the earlier project uses Mech-Vision 1.4.x or earlier, you MUST upgrade the deep learning environment to 2.0.0 or 2.1.0 (recommended) after Mech-Vision is upgraded to 1.6.1.

• If the earlier project uses Mech-Vision 1.5.x, keep deep learning environment at 2.0.0 or 2.1.0 after Mech-Vision is upgraded to 1.6.1.

Attention

Mech-Vision 1.6.0 and later no longer support the “Defect Detection” Step. This Step is available only in Mech-Vision 1.4.x with the deep learning environment 1.4.0 installed. If an earlier project still needs to use this Step, do not upgrade Mech-Vision.

Extra Python Library Should Be Installed under the “python” Directory

Mech-Vision 1.6.0 has integrated the Python 3.6.8 environment. If an extra Python library is required by the project, you must install it under the “python” directory of the Mech-Vision software.

Mech-Viz Upgrade Notes

Mech-Viz 1.6.1 has streamlined the built-in robot library. If a project uses the robot model that is not included in the built-in robot library, please download the robot model package from Robot Model Package and import it into the software, and then reload the project.

In Mech-Viz 1.6.1, the move and palletizing Tasks with the index will not automatically have the index reset when the index value exceeds the limit. Please check the project and connect “reset_task” in the proper part of the project to ensure the project can run as expected.

Mech-Viz 1.6.1 supports cylinder workobjects. If the vision result of the Mech-Vision project includes the object dimensions, you need to add the “object_type” label to the “Procedure Out” Step to notify Mech-Viz of the object type, either a cuboid, cylinder_vertical, or cylinder_horizontal.

In Mech-Viz 1.6.1, point cloud collision detection is not supported for picked cylinder workobjects. In addition, generation of object models based on received vision information and removing point cloud of picked object in the “visual_move” Task are not supported for cylinder workobjects to pick.

Mech-Center Upgrade Notes

Full-Control Program Reloading Required for ABB/FANUC/Kawasaki/KUKA

Mech-Center 1.6.1 has upgraded the full-control function. Therefore, after upgrading Mech-Center to 1.6.1, you need to reload the new full-control program for ABB/FANUC/Kawasaki/KUKA robots.

Attention

Before reloading the full-control program, please back up the current program files according to the instructions for each robot in Full-Control Program.

Usage Notes for New TM Full-Control Program

Mech-Center 1.6.1 provides a new full-control program for TM robots. The new full-control program supports controller 1.84 and higher.

If your project needs to use the new full-control program, upgrade your robot system to 1.84 or higher and reload the full-control program. If the robot in your project cannot be upgraded to the required version, please do not upgrade Mech-Center to 1.6.1.

Other Notes

Mech-Center 1.6.1 must be used with Mech-Viz 1.6.0 or later. A version mismatch may result in issues such as failing to obtain robot parameters.

If the standard interface command “Get DO List” has been used, you need to change the receiver of the “set_do_list” Task to “StandardInterface” in the Mech-Viz project.

Mech-Center 1.6.0 has built-in Python 3.6.5 environment. If extra Python libraries are required by an Adapter project, they must be installed under the “python” directory of Mech-Center.

Attention

After Mech-Center is upgraded to 1.6.1, Python libraries that have been installed on Mech-Vision 1.6.0 should be installed again.

1.6.0 Upgrade Notes

This section introduces the upgrade notes for Mech-Vision 1.6.0, Mech-Viz 1.6.0, and Mech-Center 1.6.0.

Mech-Vision Upgrade Notes

Before Mech-Vision 1.6.0, inference can be carried out only when a proper deep learning environment is installed. While Mech-Vision 1.6.0 has built-in inference models, inference can be carried out without installation of any deep learning environment.

On the industrial PCs or PCs without the deep learning environment, Mech-Vision 1.6.0 supports only the following deep learning Steps:

• Deep Learning Model Package Inference (DLK 2.2.0+): supports the models trained by Mech-DLK 2.2.0 or later

• Instance Segmentation (CPU): supports the models trained by Mech-DLK 2.2.0 or later

If an earlier project has already used deep learning Steps (other than above mentioned ones), you need to pay attention to the following notes to ensure that the earlier project can run normally after Mech-Vision is upgraded to 1.6.0:

• If the earlier project uses Mech-Vision 1.4.x or earlier, you MUST upgrade the deep learning environment to 2.0.0 or 2.1.0 (recommended) after Mech-Vision is upgraded to 1.6.0.

• If the earlier project uses Mech-Vision 1.5.x, keep deep learning environment at 2.0.0 or 2.1.0 after Mech-Vision is upgraded to 1.6.0.

Attention

Mech-Vision 1.6.0 no longer supports the “Defect Detection” Step. This Step is available only in Mech-Vision 1.4.x with the deep learning environment 1.4.0 installed. If an earlier project still needs to use this Step, do not upgrade Mech-Vision.

In addition, Mech-Vision 1.6.0 has integrated the Python 3.6.8 environment to support the “Calc Results by Python” Step. If the Python script called by this Step has used an extra Python library, you must install it under the “python” directory of the Mech-Vision software.

Mech-Viz Upgrade Notes

Mech-Viz 1.6.0 has streamlined the built-in robot library. If a project uses the robot model that is not included in the built-in robot library, please download the robot model package from Robot Model Package and import it into the software, and then reload the project.

Mech-Viz 1.6.0 does not support collision detection involving OBJ models composed of non-convex polyhedra. Collision detection of related models in earlier-version projects may fail after upgrade.

In Mech-Viz 1.6.0, the move and palletizing Tasks with the index will not automatically have the index reset when the index value exceeds the limit. Please check the project and connect “reset_task” in the proper part of the project to ensure the project can run as expected.

Mech-Viz 1.6.0 supports cylinder workobjects. If the vision result of the Mech-Vision project includes the object dimensions, you need to add the “object_type” label to the “Procedure Out” Step to notify Mech-Viz of the object type, either a cuboid, cylinder_vertical, or cylinder_horizontal.

In Mech-Viz 1.6.0, point cloud collision detection is not supported for picked cylinder workobjects. In addition, generation of object models based on received vision information and removing point cloud of picked object in the “visual_move” Task are not supported for cylinder workobjects to pick.

Mech-Center Upgrade Notes

Mech-Center 1.6.0 must be used with Mech-Viz 1.6.0 or later. A version mismatch may result in issues such as failing to obtain robot parameters.

If the standard interface command “Get DO List” has been used, you need to change the receiver of the “set_do_list” Task to “StandardInterface” in the Mech-Viz project.

Mech-Center 1.6.0 has built-in Python 3.6.5 environment. If extra Python libraries are required by an Adapter project, they must be installed under the “python” directory of Mech-Center.

In addition, Mech-Center 1.6.0 has changed the usage of some standard interface commands. For details, refer to Mech-Center 1.6.0 Release Notes .