Release Notes

Mech-DLK 2.5.2 Release Notes

This section introduces the new features, improvements, and resolved issues of Mech-DLK 2.5.2.

New Features

Added the Pre-labeling Feature

After you validate a model, you can import new image data to the current module and use the pre-labeling feature to perform auto-labeling based on this model. You can fine-tune the results after labeling is completed.

  • Supported modules: Instance Segmentation, Object Detection, Classification, Defect Segmentation, and Text Detection.

  • Prerequisites: The pre-labeling feature is available only when the current module contains validated models.

  • The pre-labeling feature can be performed only on the following three types of data:

    • Unlabeled data

    • Automatically labeled data (images with a yellow triangle at the front of the image number)

    • Manually fine-tuned data after automatic labeling (images with a yellow triangle at the front of the image number)

  • Three methods to perform the pre-labeling operation:

    • The Pre-labeling Tool in the labeling toolbar

    • The Pre-label button in the upper part of the image list

    • The Pre-label option in the right-click menu

    For more information, see Pre-labeling.

Added the Image Tag Drop Down List

The Image tag drop down list is now available when you use a cascading module and import data from the previous module. You can select the tag that you need to import corresponding images. For more information, see Use Cascaded Modules.

Added the Label Merging Feature

Mech-DLK 2.5.2 added the label merging feature. When you select a label, you can right-click the label and select Merge Into to change the current data type to another type.

  • Supported modules: Classification, Instance Segmentation, and Object Detection.


Supported the Import of Datasets in the COCO Format

Some modules in Mech-DLK 2.5.2 allow you to import datasets in the COCO format.

  • Supported modules: Instance Segmentation and Object Detection.

Integrated the Data Import and Export Features

In Mech-DLK 2.5.2, the Import button in the upper-left corner of the user interface is replaced by an Import/Export button, which integrated the data import and export features.

You can perform the import and export operations more easily.

Optimized the Data Export Settings

Mech-DLK 2.5.2 optimized the data export settings (also applicable to cascading modules):

  • Export all data from the current module: In the upper-left corner of the user interface, click the Import/Export button, then select Export Dataset.

  • Export part of the data from the current module: In the image list, select one or more images, then right-click and select Export Dataset.

Resolved Issues

The following issues have been resolved in Mech-DLK 2.5.2:

  • Fixed the issue that the verification results were also displayed in single image masks or global masks configured in the Defect Segmentation module.

  • Fixed the issue that global masks configured in the Defect Segmentation module did not scale proportionally to the size of images.

  • Fixed the software crash caused by unusual operations when using the Rectangle Tool.

Mech-DLK 2.5.1 Release Notes

Mech-DLK 2.5.1 improved the Smart Labeling Tool to enhance the labeling efficiency.

Mech-DLK 2.5.0 Release Notes

This section introduces the new features and improvements of Mech-DLK 2.5.0.

New Features

Added Text Detection and Text Recognition Modules

Mech-DLK 2.5.0 now features optical character recognition (OCR) that is realized by cascading the following two algorithm modules:

  • Text Detection: the first module, which is used to detect a single line or multiple lines of text in an image.

  • Text Recognition: the second module, which is used to recognize the characters in the text area.

This feature is usually employed to detect and recognize tiny characters on electronic components, vehicle-related data against complex backgrounds, and serial numbers and lot numbers displayed on the packaging.

Added the Unsupervised Segmentation Module

Mech-DLK 2.5.0 now features the Unsupervised Segmentation module, which only requires OK images in the training set to achieve pixel-level detection of all known and unknown defects. The trained model can be used to judge whether the image of an object is OK, NG, or Unknown on the basis of set thresholds. Moreover, heat maps are available to show the possible areas with defects.

This module is usually used for industrial quality inspection in scenarios where

  • obtaining OK images is easy while getting NG images is difficult, and

  • defect types are uncertain.

Added the Free Rectangle Tool

In the Text Detection module, the Free Rectangle Tool can be used to draw a rectangle around the text area. It is recommended to select this tool for rectangular text areas.

Added Text Recognition Tool

In the Text Recognition module, the Text Recognition Tool can be used to define the text recognition range and automatically generate the recognition result.

Added OK/NG Labels

Mech-DLK 2.5.0 now allows to directly label images as OK or NG with the OK/NG Labels. Select an image from the image list and click the OK Label or NG Label on the toolbar. The image will be labeled as OK or NG.

  • Both OK Label and NG Label are available in the Unsupervised Segmentation module.

  • OK Label is available in the Defect Segmentation module.


Support for NVIDIA GeForce RTX 40 Series GPUs

Mech-DLK 2.5.0 now fully supports NVIDIA GeForce RTX 40 Series GPUs for model training and validation.

Improved the Floating Window of Keyboard Shortcuts

In Mech-DLK 2.5.0, the floating window of keyboard shortcuts displays the shortcuts by modules for easy and fast search.

Release Notes of Previous Versions

Click here to view Mech-DLK 2.4.x release notes

Mech-DLK 2.4.2 Release Notes

Added region-specific license control. Click Help  About to view the details.

Mech-DLK 2.4.1 Release Notes

New Features

  • Added the Cascade Mode

    Mech-DLK 2.4.1 has added a brand-new cascade mode of modules, which enables the combination of modules to solve deep learning problems in complex scenarios. It should be noted that when the Fast Positioning module is involved, it must be the first module. For example, when you need to detect the positions of defects and classify these defects, you can first add a Defect Segmentation module and then add a Classification module. In addition, when importing data from the previous module, you can select images according to actual needs and configure the import of these images in the Import window.

  • Added the Training Center

    The Training Center supports the training of models in a queue, which is suitable for scenarios requiring the training of multiple models. With the Training Center, the software can train models in sequence, with no need for manual clicking of Train repeatedly, which can save a huge amount of time.

  • Added the Mask Type of Mask Globally and Supported Custom Mask Fill

    In the Defect Segmentation module, when you select the Mask Polygon Tool, the mask type can be Mask single image or Mask globally. In addition, you can customize the mask color.

    • Mask single image: The mask is only displayed in the current image. It is only valid in training.

    • Mask globally: After a mask is drawn in the current image, the mask will be displayed in all images. The mask is valid in both training and validation.

  • Added the Floating Window of Keyboard Shortcuts

    Click keyboard shortcut keyboard in the lower-right corner of the selection region to open the window of keyboard shortcuts.

  • Added Auxiliary Labeling Lines for Rectangle Tool

    In the Instance Segmentation and Object Detection modules, auxiliary labeling lines are added for the Rectangle Tool to assist rectangular selection on images.

  • Added the Display and Filtering of Validation Result Confidence

    In the Instance Segmentation and Object Detection modules, a confidence filtering function is added for validation results. You can adjust the confidence to filter the validation results and then evaluate the accuracy of models.


  • Optimized the Classification Module

    The Classification module is optimized, which leads to faster training convergence and a growth rate of 20% in accuracy under complex scenarios.

  • Optimized Mech-DLK SDK

    Mech-DLK SDK is more stable and easier to use after being restructured. Mech-DLK SDK supports the inference based on the cascaded modules and the switching between different operating hardware. In addition, it provides richer samples for reference.

  • Optimized the Setting of Defect Determination Rule

    The defect determination rule in the Defect Segmentation module is optimized. Click here to view the details.

  • Enabled the Setting of Translation in the Fast Positioning Module

    In the Image Adjustment window of the Fast Positioning module, you can translate the image along the X and Y axes. After training, images with objects in specified positions and orientations will be generated, which meet the requirements of more application scenarios.

  • Optimized the Template Tool

    In the Instance Segmentation and Object Detection modules, after selecting the Template Tool, press and hold the Shift key and scroll the mouse wheel to adjust the angle of the template. You can also set the Rotation angle to achieve the same purpose.

Click here to view the Mech-DLK 2.3.0 release notes

Mech-DLK 2.3.0 Release Notes

  • Graphics Card Driver Requirement

    Before using Mech-DLK 2.3.0, please upgrade the graphics card driver to 472.50 or above.

  • Improved the Training Speed

    Optimized the algorithms, and thus significantly improved the speed of model training. Only the optimal model is saved during training, and the training cannot be stopped halfway.

  • Added the Smart Labeling Tool

    For modules including Defect Segmentation, Instance Segmentation, and Object Detection, you can do smart labeling by selecting the Smart Labeling Tool, clicking the objects to be labeled, right-clicking to undo the redundant selection, and pressing the Enter key to complete the labeling.

  • Added the Function of Adding/Removing Vertices for the Polygon Tool

    For the Instance Segmentation and Object Detection modules, after labeling with the Polygon Tool, if the selection needs to be modified, you can left-click the line segment between two vertices to add a vertex, or right-click a vertex to remove it.

  • Added the Template Tool

    For the Instance Segmentation and Object Detection modules, you can use the Template Tool to set the selection as a template. The template can be applied by simply clicking the images. It is suitable for scenarios where there are multiple neatly arranged objects of the same type in an image, and it improves labeling efficiency.

  • Added the Function of Preview by Zooming

    Support previewing full images and cropped cell images.

  • Optimized the Grid Cutting Tool

    Optimized the Grid Cutting Tool. After cutting the image by the grid, you can select a cell image by checking the box in the upper left corner of the cell image, and you can preview the image by clicking on the button in the upper right corner of the cell.

  • Optimized the Data Filtering Mechanism

    Added options for filtering results: “Correct results”, “Wrong results”, “False negative”, and “False positive”. Added options for filtering data types: “Labeled as OK” and “Labeled as NG”.

  • Built-in Deep Learning Environment

    The deep learning environment is built into the software Mech-DLK, and the models can be trained without a separately installed environment.

Click here to view Mech-DLK 2.2.1 release notes

Mech-DLK 2.2.1 Release Notes

  • Added the Function of Showing the Class Activation Maps for Module Classification

    After the model is trained, click Generate CAM. The class activation maps show the weights of the features in the form of heat maps; the model classifies an image into its class according to these features. Image regions with warmer colors have higher weights for classifying the image into its class.

  • Supported Validation and Export of CPU Models

    • Classification and Object Detection: After training is completed, select the deployment device as CPU or GPU before exporting the model.

    • Instance Segmentation: Before training the model, set the training parameters. When exporting a model, select the deployment device as CPU/GPU:

      • CPU lightweight model: Before training the model, set the training parameter Model type to Lite (better with CPU deployment). When exporting the model for deployment, set Deployment device to CPU or GPU.

      • GPU standard model: Before training the model, set the training parameter Model type to Normal (better with GPU deployment). When exporting the model for deployment, set Deployment device to GPU.