Features of previous software versions

This post only briefly introduces the key features of some previous software versions. If you need to download relevant material, please proceed to Downloads.

Mech-Mind Software Suite



  • “Medicine Boxes” scenario is added to the “Predict Pick Points V2” Step
  • “Deep Learning Model Package Management Tool” supports “Efficiency Configuration” of the model package inference
  • New Step “Deep Learning Model Package Inference”
  • New Step “Deep Learning Result Parser”


  • Simulating all planned waypoints
  • Support more Euler angle conventions


  • Standard Interface supports the “Mitsubishi MELSEC Protocol”
  • Standard Interface supports “UDP”
  • Standard Interface supports integration with ROKAE industrial robots, HYUNDAI industrial robots, and NACHI industrial robots
  • New URCap Plugin 1.5.0 for “PolyScope 6 System”
  • Standard Interface supports communication with AB PLC based on EtherNet/IP



  • One-stop software user experience
  • Advanced component “3D Workpiece Recognition”
  • Solution template “General Workpiece Picking”
  • Managing all project data as “solution”
  • Solution library with more than 70 cases


  • New tree-structured “Project Resources” interface
  • Auto-generation of pallet pattern for single-size cartons
  • Refactoring depalletizing algorithms
  • Enlarging the online robot library to include more than 600 robot models


  • Supporting communication configuration via Mech-Vision
  • Optimizing Standard Interface
  • Master-control support for ESTUN and AUBO robots
  • Optimizing the loading of Adapter projects

Mech-Eye SDK


  • Improved point cloud quality: the Point Cloud Processing parameters have been redesigned and can effectively remove outliers and noises while presevering the object edge information in the point cloud, improving point cloud quality.
  • Enhanced the functionality of interfaces:
    • Parameters used for hand-eye calibration are added to the GenICam interface and Mech-Eye API, and corresponding HALCON sample programs are provided. You can now perform hand-eye calibration directly with the interfaces.
    • Parameters for obtaining multi-part payload type data are added to the GenICam interface, and corresponding sample programs are provided. You can now easily obtain textured point clouds and depth maps that contain only the Z values.
    • A firmware upgrade tool is added. You can now upgrade the camera firmware without Mech-Eye Viewer.


  • Compliance with the machine vision industrial standards GenICam 3.0 and GigE Vision 2.1.
  • New methods for mapping 2D images to depth maps, with corresponding sample programs.



  • Flexible model cascading: Now, you can combine modules to solve deep learning problems in complex scenarios.
  • Queued training: Models can queue up for auto training in the Training Center. With this function, the software can train models automatically in turn, so you no longer need to stay in front of the computer to train the models one by one.
  • More features in Mech-DLK SDK: Mech-DLK SDK now supports the inference of cascaded models and the switching among different operation hardware. What’s more, more samples are ready this time.
  • More intuitve defect determination rules.
  • Faster and more accurate “Classification” module: The improved Classification module features faster convergence and an increase of 20% in accuracy under complex scenarios.