Labgym

1. Implemented the analysis pipeline for complex interactive behaviors.
2. Major improvement on analysis speed.
3. Bug fix.

1. Implemented Detector-based detection method. Now the changing background or illumination in videos is no longer a problem for LabGym.
2. Implemented data mining functional unit.
3. Implemented preprocessing functional unit.
4. Simplified code and further optimized the processing speed.
5. Implemented basic analysis for interactive behavior.

1. In previous versions, if no animal is detected in a frame, this frame will be skipped. From now on, such frames will not be skipped, and the behavioral classification and quantification will be output as 'NA's so that the raster plots and the quantification results can be perfectly aligned for every frame with other data (e.g., ephys recordings).
2. An 'uncertain level' can be added into the Categorizers for reducing the false positives in behavior classification. The Categorizer will output an ‘NA’ if the difference between probability of the highest-likely behavior and the second highest-likely behavior is less than the uncertainty level.
3. Simplified the user interface.
4. Optimized the processing speed.

1. Improved the background extraction and the tracking, making them faster and more accurate.

1. Added a version checker. If a newer version of LabGym is available, users will see a reminder when initiate the user interface.

1. Simplified the user interface and made it more self-illustrative.
2. Added an option of whether to output the distances in pixels when calculating behavior parameters. Previously all the distances were just normalized by the size of a single animal.