0.0.3

What's Changed

• fix: Failure in cyto3 flow-field calculation on CUDA by @joda01 in #200
• fix: Images with no date wer not displayed in output table by @joda01 in #200
• fix: Buffer OME to avoid multiple read by @joda01 in #200
• fix: When grouping based on filename is active show the filename instead of the coordinates by @joda01 in #200
• fix: Fixes for macOS 15 by @joda01 in #200
• feat: Added ad hoc machine learning using Random forest or ANN MLP by @joda01 in #200
• feat: Allow to disable CUDA usage by @joda01 in #200
• feat: Add pseudo color support for preview image by @joda01 in #200
• feat: Allow to manual label areas in image by @joda01 in #200
• feat: Allow single image analyzing by @joda01 in #200
• feat: Add keyboard shortcuts for channel selection and preview refresh by @joda01 in #200
• feat: Histogram settings are now saved for each image by @joda01 in #200
• feat: Performance improvements in image loading by @joda01 in #200
• feat: Added live object metrics preview by @joda01 in #200

Highlights

Machine learning support

ImageC 0.0.3 does allow pixel classifiers to be trained directly within the program.
For deep learning networks, a feed-forward artificial neural network (ANN) multi-layer perceptron (MLP) is used, which is generated with PyTorch.
This allows for fast training and execution thanks to CUDA support.

Manual annotation

ImageC 0.0.3 enables manual object annotation alongside pipeline annotation.
Both methods can be used for AI training or analysis.
One highlight is the possibility to use pipeline annotated objects for training of AI models, which can be used to fasten up the time for labeling.

Full Changelog: 0.0.2...0.0.3

What's Changed

• feat: Added ad hoc machine learning by @joda01 in #200

Full Changelog: 0.0.2...0.0.3