The Scaffolded Sound Beehive

AnneMarie Maes

OKNO - Brussels Urban Bee LabVlaamse Steenweg 66

1000 Brussels - Belgium.annemie@okno.be

Abstract

The Scaffolded Sound Beehive is an immersivemulti-media installation which provides viewers anartistic visual and audio experience of activities ina beehive. Data were recorded in urban beehivesand processed using sophisticated pattern recogni-tion, AI technologies, and sonification and com-puter graphics software. The installation includesan experiment in using Deep Learning to interpretthe activities in the hive based on sound and micro-climate recording.

1 Description of the work

I propose an installation centering around the theme of beemonitoring. The centerpiece of the installation is the top of aWarré beehive constructed using open source digital fabrica-tion and mounted on scaffolds. A Warré hive is a sustainablebeehive originally designed by the Warré around 1920. Thehive is 2.5 m high so that visitors can put their head insideit and experience a visual and auditory artistic interpretationof hive activity, making this an interactive immersive instal-lation. The Scaffolded Hive will be a very visible beautifullandmark (Figure 1).

Three further components are added:1. An 8-channel sound installation plays continuously in-

side the hive. This sound installation is based on record-ings of actual bee and environmental sounds in thebroodnest of an urban beehive installed on the roof topof the Brussels Urban Bee Laboratory for a completeseason (Fig. 2, right). It started with recordings onJune 21st 2014, the longest day/shortest night, processedusing sophisticated pattern recognition algorithms, andartificial intelligence analysis software, and edited intoa 15 minute-piece using Max/MSP by adding swirlingelectronic sound clusters to sonify the increase and de-crease of swarm activity in the hive.

2. A video shows 365 days of activity inside a real obser-vation beehive, played back at higher speed. The imageswere recorded with an infrared camera inside the hive(Fig. 2, left) and processed using pattern recognition,AI and computer graphics algorithms (Fig. 3). Theseimages give a stunning visual experience of a colony inaction.

Figure 1: Left. Physical Warré-hive. Right: 3d Model (constructedwith Blender) of the Scaffolded Warré hive. The Scaffoled Hive willbe digitally fabricated based on this design.

3. A second video shows a graphical rendering of AI anal-ysis of colony behavior combining real audio data withmeasurements of microclimate inside the hive: temper-ature, CO2 and humidity and using DEEP LEARNINGto detect patterns in the data (Fig. 4).

2 Artistic goals and production

The Scaffolded Sound Beehive is part of a series of ecologicalinstrumented beehives leading towards a fully biocompatibleintelligent beehive [Maes, 2013]. These beehives allow me tostudy the tight interaction between city honeybees and urbanecosystems, using artistic research practices and in collabora-tion with scientists. My preoccupations with bees come partlyfrom a fascination with these amazing insects: Bees exhibitvery original solutions to the challenges that social insectsface, e.g. on the level of communication and collective deci-

Figure 2: Field recordings. Left: Cameras connected to RaspberryPi computers film, process and stream the visual data recorded inthe hive. Right: Piezo sensors are inserted into the hive, between thecomb frames and in front of the entrance on the landing platform.The piezo’s are connected to a small mixing panel with amplifica-tion.

sion making. They are an endless source of visually stunningimages and sounds and their remarkable collective behaviorprovides inspiration and metaphors for the functioning of hu-man society. But in many industrialised nations, bee coloniesare now threatened. So I work also towards an improvementof the environment of bees with the creation of urban gardensand guerilla planting and I use bees as bio-indicators to makecitizens aware of the increasingly negative effects of our lifestyles and methods of industrial production.

Figure 3: Top. Extract from visually observed bee activity. Bottom:processing and tracking of bee movements.

3 Relationship to AI

The Scaffolded Sound Beehive uses a broad range of pat-tern recognition and AI techniques. As a special contribu-tion to IJCAI, a project was carried out to use Deep Learningtechniques using massive data from sound and microclimatein order to get an interpretation of hive activity. The datawere transformed into a multi-dimensional feature space, andpresented to a Deep Learning network to extract categoriescapturing patterns of activity through unsupervised learning.An original visualisation was developed (see Figure 4) whichrepresents categories as attractors in a projected 3d space,with the size of the attractor reflecting the strength of the cat-egory.

Figure 4: Novel visualisation of results of Deep Learning. Thespheres represent learned categories. The white dots are hypotheti-cal bees spatialized for how well they fit with the categories.

4 Context of the work

The Brussels Urban Bee Lab [BUBL] is an independent in-ternational collective of artists, scientists, beekeepers, tech-nicians and creative people. It was founded by AnneMarieMaes in 2009. BUBL uses artistic, scientific and technolog-ical research approaches to tackle challenges related to sus-tainability and the monitoring and survival of city honeybees.An extensive rooftop garden in the centre of Brussels hoststhe headquarters of the BUBL laboratory. It is giving placeto several experimental beehives and instrumentation equip-ment. From this laboratory, data is continuously being broad-cast through streaming technology.

The Brussels Urban Bee Lab operates mainly through artis-tic installations and workshops. The group has produced sev-eral very original beehives and produced art works shown ininternational exhibitions. The installations explore highly ex-perimental technologies, such as microbial fuel cells, digitaland organic fabrication using OpenStructures, biomimicry,spatialised sonification, web-based continuous data stream-ing, data mining based on Artificial Intelligence, organic elec-tronics, solar energy for powering low-energy computing, etc.

5 Requirements

The hive will be shipped to Buenos Aires. The installation ofthe hive is straightforward. 8 m2 is the minimal space. Onlyelectricity is needed. Monitors for the videos can be hired ortablets can be brought. The sound system is embedded in thehive.

6 Acknowledgement

Members of the Brussels Urban Bee Lab that helped to realizethe current installation include Balthazar de Tonnac, YacineSebti, Vincent Malstaf, Lennert De Taye, Billy Bultheel, JoeriBultheel Bob Mottè. Funding was received from the FlemischMinistry of Culture and the Flemish Community Commis-sion of the Brussels-Capital Region, as well as the CultureProgram of the European Community (ALOTOF project).

References

[Maes, 2013] AnneMarie Maes. The Transparent Beehive.OKNO Publications, Brussels, 2013.