Software for honeybee trials using AI
Honeybee trials can considerably profit from using software tools and digital photography for their evaluation. The use of photographs reduces the impact on the colonies to a minimum and with modern software technology based on AI (artificial intelligence) and image recognition algorithms the evaluation is much faster and more reliable than any previous methods. We have developed several products for your evaluations:
HoneybeeComplete (colony assessments & brood trials)
HoneybeeComplete (for MS Windows) is our prime product. It includes all functions of IndiCounter and Honeybee Brood Logger (discontinued) plus some additional features. With HoneybeeComplete you can do honeybee brood trials (e.g. according to OECD 75 or Oomen et al. 1992) and colony assessments. Instead of relying on colony estimation methods (e.g. Liebefelder) you can profit from counting honeybees or capped cells directly with our auto-recognition addons. Also cell content (e.g. eggs, larvae, pollen, etc.) can be detected automatically with a high accuracy (acceptable image quality provided, the computer can only see what your eye can also see). Here is an overview of the accuracy of auto-recognition addons (tested using 10 images with 36 MP resolution):
| Items | Accuracy (predicted vs. observed) |
|---|---|
| Carnica bees (adults) | 99.7% |
| Buckfast bees (adults) | 99.6% |
| Detection of cells (location of cells) | 98.9% |
| Detection cell content (eggs, young larvae, old larvae, capped brood, pollen, nectar, honey) | 95% |
In addition, recognition can be restricted to user defined areas of images. If you still need a better performance, we can train the software for your particular camera setup. Colony assessments can be based either on the number of bees (or capped cells) or on coverage. For brood trials addons are available for the automatic detection of eggs, young larvae, old larvae and pupae.
IndiCounter (colony assessments)
IndiCounter for colony assessmentsIndiCounter is a software for MS Windows for honeybee colony assessments or other biological trials which are based on counting individuals. You can measure colony size much more reliably by counting individuals or capped cells with our automatic image recognition addons. Good image quality provided the software can recognise close to 100% of adult bees and capped cells (see table above). In addition, recognition can be restricted to user defined areas of images. If you still need a better performance, we can train the software for your particular camera setup. No other methods, including the Lieberfelder estimation or methods based on colour similarity, reach such accurate results.
Feature matrix
| Feature | HoneybeeComplete | Honeybee Brood Logger* | IndiCounter |
|---|---|---|---|
| Automatic adjustment of images (no specific photo-box required) | Yes | Yes | Yes |
| Brood evaluation according to OECD 75 | Yes | Yes | – |
| Brood evaluation according to Oomen et al. 1992 | Yes | Yes | – |
| Continuous brood evaluation (long-term monitoring) | Yes | – | – |
| Image gallery displaying development per cell | Yes | Yes | – |
| Honeybee colony assessments | Yes | – | Yes |
| Direct printing of results with a watermark (GLP) | Yes | – | Yes |
| Evaluation of multiple photos/combs | Yes | Yes | Yes |
| Tabular and graphical ouput of results | Yes | Yes | Yes |
| Suitable for auto-recognition addons (e.g. adult bees, eggs, young larvae, old larvae, etc.) | Yes | – | Yes |
Version history
HoneybeeComplete Version 7.2
- Improved recognition of AI modules:
- Carnica bees: recognition accuracy (predicted vs. manual counts) of 99.7%
- Buckfast bees: recognition accuracy (predicted vs. manual counts) of 99.6%
- Recognition of cells (for detecting cells at first BFD): recognition accuracy 98.9%
- Recognition of cell content (average of larvae, eggs, pollen, etc.): 95%
- Accuracy was tested with 36 MP images and may differ for different resolutions and setups.
- Big fixes and minor improvements
HoneybeeComplete Version 7.1
- Audit trail Addon
- Big fixes and minor improvements
HoneybeeComplete Version 7.0
- New AI module for recognition of bees and cell content
- 3D view of the hive
- New options to adjust cell positions separately for any BFD
- Big fixes and minor improvements
HoneybeeComplete Version 6.0
- Additional option to display of brood stages as coloured circles around cells.
- Custom colours for the display of brood stages.
- Additional features for marking cells.
- Display of cells and brood stages when double clicking in the result tables (OECD 75 table and full evaluation table).
- Option to edit cell content in the result tables (OECD 75 table and full evaluation table).
- Bug fixes and minor improvements.
HoneybeeComplete Version 5.4
- Minor improvements.
HoneybeeComplete Version 5.3
- Bug removals.
HoneybeeComplete Version 5.2
- Minor improvements.
HoneybeeComplete Version 5.1
- Minor improvements.
HoneybeeComplete Version 5.0
- Completely rewritten auto-recognition module, making the recognition more stable (several memory problems solved).
- Increased recognition of addons for adults bees, old larvae, young larvae, eggs.
- Inclusion of a new pollen recognition module (separate addon).
- New option to limit the presence of brood or pollen to a specific part of the image (this replaces an option where only capped brood and capped nectar could be distinguished).
- Removal of a few bugs.
HoneybeeComplete Version 4.2
- Custom brood development schemes can now be used. Minor other improvements.
HoneybeeComplete Version 4.1
- Removal of a bug occuring when loading files.
HoneybeeComplete Version 4.0
- Addition of a new evaluation method for monitoring the frequency of brood stages over a longer time period (continuous hive evaluation).
- Automatic recognition of nectar cells and capped honey cells. No separate recognition-addon is needed, this works in combination with addons for eggs, larvae, and capped brood.
- Addition of an easy option for manual distinction of capped honey cells from capped brood cells using a polygon method.
HoneybeeComplete Version 3.1
- Optimisations for displaying cell numbers and brood stage letters (clearer especially on dark background). Options to accept further deviations of larval and pupae development (e.g. 12 d pupae stage with 4 d larval stages) as normal brood development. Minor other improvements.
HoneybeeComplete Version 3.0
- Capability to use autorecognition addons for automatic marking of cells and for recognition of empty cells, eggs, young larvae, old larvae and capped cells for evaluation of brood trials
HoneybeeComplete Version 2.1
- Improved memory management
HoneybeeComplete Version 2.0
- New evaluation module for coverage based on polygon areas
- Option for enhanced auto-recognition for low image resolution
HoneybeeComplete Version 1.3
- Optional restriction of auto-recognition to user defined areas of images
- Optional recognition in all images of a project
- Optional import of multiple photos at once
HoneybeeComplete Version 1.2
- Added option for reduced memory usage for processing large amounts of images
- Added compatibility for auto-recognition addons
- Added evaluation based on coverage
HoneybeeComplete Version 1.1
- Added watermark which disappears when copying prints (for GLP studies)
- Minor improvements and bug fixes
HoneybeeComplete Version 7.2
- Improved recognition of AI modules:
- Carnica bees: recognition accuracy (predicted vs. manual counts) of 99.7%
- Buckfast bees: recognition accuracy (predicted vs. manual counts) of 99.6%
- Recognition of cells (for detecting cells at first BFD): recognition accuracy 98.9%
- Recognition of cell content (average of larvae, eggs, pollen, etc.): 95%
- Accuracy was tested with 36 MP images and may differ for different resolutions and setups.
- Big fixes and minor improvements
Indicounter Version 5.1
- Audit trail Addon
- Big fixes and minor improvements
Indicounter Version 5.0
- New AI module for recognition of items
- Big fixes and minor improvements
Indicounter Version 4.0
- Completely rewritten auto-recognition module, making the recognition more stable (several memory problems solved).
- Increased recognition of addons for adults bees, old larvae, young larvae, eggs.
- Inclusion of a new pollen recognition module (separate addon).
- Removal of a few bugs.
IndiCounter Version 3.1
- Optimisations for displaying cell numbers and brood stage letters (clearer especially on dark background). Minor other improvements.
Indicounter Version 3.0
- New evaluation module for coverage based on polygon areas
- Option for enhanced auto-recognition for low image resolution
- Improved memory management
IndiCounter Version 2.4
- Optional restriction of auto-recognition to user defined areas of images
- Optional recognition in all images of a project
- Optionl import of multiple photos at once
IndiCounter Version 2.3
- Added option for reduced memory usage for processing large amounts of images
IndiCounter Version 2.2
- Added compatibility for auto-recognition addons
- Added evaluation based on coverage
IndiCounter Version 2.1
- Added watermark which disappears when copying prints (for GLP studies)
IndiCounter Version 2.0
- New graphical interface
- Minor improvements and bug fixes
IndiCounter Version 1.1
- Minor improvements and bug fixes
Related links
Photography for Honeybee Brood Trials
High quality photography is a prerequisite for high throughput analysis of brood development and for successful automated counting of adult bees.
Software for video assessment of forager losses in bees
VideoCounter is software for automated counting of honeybees or bumblebees leaving and entering their hives.
Low variability and uncertainty honeybee trials (GLP)
In order to meet protection goals defined by EFSA low variability and low uncertainty honeybee field or semi-field trials are needed.
