Version Compatibility#

  • 0.9.0

    • 0.8.2: L1 - Fully compatible.

      • AssemblyVerification: L4 - Add a new option to select the image pages to be processed in this tool. Added property node mask as input of configurator node statistician to filtered the VFLOW:Without-Training regions when calculating statistical metrics.

      • Classification: L3 - Add a new option to select the image pages to be processed in this tool.

      • Detection: L4 - Add a new option in training parameter to let to enable users to select whether view masks are involved in model training or not. And add a new option to select the image pages to be processed in this tool. Fix the issue where candidate boxes in overlapping areas between multiple views of the original image cannot be filtered by setting the overlap ratio. Added property node mask as input of configurator node statistician to filtered the VFLOW:Without-Training regions when calculating statistical metrics.

      • ELClassification: L3 - Add a new option to select the image pages to be processed in this tool.

      • ELOCR: L4 - Add a new option to select the image pages to be processed in this tool. Added property node mask as input of configurator node statistician to filtered the VFLOW:Without-Training regions when calculating statistical metrics.

      • ELUnsuperClassification: L3 - Add a new option to select the image pages to be processed in this tool.

      • ELUnsuperSegmentation: L4 - Upgrade training and inference performance. Note that you should retrain your model after upgrading the tool.

      • GeometrySearch: L3 - Add a new option to select the image pages to be processed in this tool. Add a method to deduplicate results if views overlap.

      • Input: L4 - A new property node named image_user_data has been added to store the thumbnail image and other user data of the sample. After the upgrade, the old version of the property node image_info will no longer be able to store the thumbnail image, and the thumbnail image in old version projects will be transferred to property node image_user_data.

      • Location: L4 - Add a new option to select the image pages to be processed in this tool. Results covered by view masks will be filtered out. Added property node mask as input of configurator node statistician to filtered the VFLOW:Without-Training regions when calculating statistical metrics. Add a method to deduplicate results if views overlap.

      • OCR: L4 - Add a new option to select the image pages to be processed in this tool. Added property node mask as input of configurator node statistician to filtered the VFLOW:Without-Training regions when calculating statistical metrics.

      • Segmentation: L4 - Added property node mask as input of configurator node statistician to filtered the VFLOW:Without-Training regions when calculating statistical metrics. Add a new option in training parameter to let to enable users to select whether view masks are involved in model training or not. And add a new option to select the image pages to be processed in this tool.

      • UnsuperClassification: L3 - Add a new option to select the image pages to be processed in this tool.

      • UnsuperSegmentation: L4 - Added property node mask as input of configurator node statistician to filtered the VFLOW:Without-Training regions when calculating statistical metrics. Add a new option in training parameter to let to enable users to select whether view masks are involved in model training or not. Added contrastive mode. This mode is only applicable to the multi-visual image composed of two images, and allows adding different models of the same product for training to detect all model product defects. And add a new option to select the image pages to be processed int the tool.

      • ViewTransformer: L4 - Add feature to support transmit view masks from the Input tool.