EcoAssist v5.1

Moved PyQt5 dependency to PySide6 for Unix systems to avoid HomeBrew dependency and extra install.

Full Changelog: v.5.0...v5.1

EcoAssist v5.0

• new window: simple mode
• saves user settings
• incorporated Namibian species model
• automatic model download
• switch between virtual environments

What's Changed

• Update to v5.0 by @PetervanLunteren in #34

New Contributors

• @PetervanLunteren made their first contribution in #34

Full Changelog: v4.4...v.5.0

EcoAssist v4.4

In previous versions of EcoAssist (v3.0 > v4.3) it was possible to train your own object detection models based on MegaDetector to detect your target species. Although this did work, it wasn't the best approach to develop a species recognition model. It required lots of training data, processing power, time, electricity and wasn't very accurate. Advancing insights revealed that better results can be obtained by using a object classification model to be used in conjunction with the results of MegaDetector. The animals will then be located by MegaDetector, and further classified by your custom model. EcoAssist > v4.2 does support the deployment of a classification model to be used in conjunction with MegaDetector, but training such a model is more complicated and hasn't been incorporated into EcoAssist (yet).

Further:

• new logo
• fixed scroll issues
• moved documentation to https://addaxdatascience.com/ecoassist/

Full Changelog: v4.3...v4.4

EcoAssist v4.3

Full Changelog: v4.2...v4.3

EcoAssist v4.2

• Annotation in xml to be able to store more information
• human in the loop feature
• minor fixes

Full Changelog: v4.1...v4.2

EcoAssist v4.1

Added human-in-the-loop functionality to verify detections and create training data. Possibility to select an image subset based on specific classes, confidence ranges, and selection methods.

What's Changed

• Automatic localization of git and conda during Windows installation by @BobbyCephy in #11
• Copied Exif data from source image during post processing by @ehallein in #14

New Contributors

• @BobbyCephy made their first contribution in #11

Full Changelog: v4.0...v4.1

EcoAssist v4.0

Added functionality to train object detection models and fixed bugs.

What's Changed

• Fixed a CUDA out of memory bug by @ehallein in #7

Full Changelog: v3.0...v4.0

EcoAssist v3.0

• New interface
• Postprocessing features work with custom classes
• Accommodates both new and old yolov5 models
• Separate confidence thresholds per feature
• New feature to sort data based on confidence

EcoAssist v2.2

• All features work with relative path
• New option to copy or move files when separated
• Fixed bug regarding faulty checkpoint frequency and sample rate
• Lowered the default threshold for images to 0.1 and videos to 0.2
• Improved information text about threshold values and further processing in Timelapse
• Added message box after processing if created .xml files

EcoAssist v2.1

Bug fixes and thoroughly tested.

Features:

• Use either MDv5a or MDv5b to tag animals, persons and vehicles in both images and video's
• Filter out empty images, people, vehicles or animals
• Review and edit annotations using the open-source annotation software labelImg
• Create input file for further processing in Timelapse
• Export .xml label files in Pascal VOC format for further model training
• Manipulate data by drawing boxes or cropping detections
• Use custom models trained from MDv5 using transfer learning