Avanade Intelligent Garden at the Chelsea Flower Show

Inviting visitors to interact with the plants at the 2025 Chelsea Flower Show in London, UK, proved a positive showcase for AI and Raspberry Pi 5.

Keen gardener King Charles III famously promoted the idea of talking to plants in a 1986 episode of children’s TV show Blue Peter. This same idea could be found at this year’s Chelsea Flower Show, where technology company Avanade embedded sensors in trees and used an app to prompt conversations about each tree’s health. 

A Pollinator Camera installed on the pavilion roof at the Chelsea Flower Show recorded insect visitors; Raspberry Pi 5 and a Hailo AI accelerator automatically recognised and recorded photos of bees and butterflies

Alongside this, Avanade’s Pete Gallagher created a bespoke Pollinator Camera, based on Raspberry Pi 5 and the Hailo AI accelerator, to photograph and detect the number of insects visiting the show garden each day. The Intelligent Garden won a gold award for Best Construction at the Chelsea Flower Show. 

Trees, please

Schoolchildren and conservation charities often employ some form of AI to record wild visitors, using hidden cameras and Raspberry Pis to identify them. Avanade decided to bring a similar form of citizen science to people who may be unacquainted with how AI — or, more properly, machine learning — could be used to enhance their own gardens.

Pete in the Intelligent Garden with the Pollinator Camera hardware

To this end, Avanade approached renowned gardeners Tom Massey and Je Ahn to design an appealing garden with plenty of different species of trees, as well as plants that would attract insects and butterflies. The Times journalist Lucy Bannerman described the Intelligent Garden as a “garden that tells people how it is feeling”. Given the blazing heat in London in late May, it mostly reported (via an auto-generated text message) being pretty thirsty!

Avanade’s Helen Woodfield explains that plenty of trees are being planted in the UK, with leaf cover helping to cool streets, remove pollutants, and provide much-needed shade. However, all the stats say that street trees in cities are really suffering; 50% die within ten years, wasting money and the goodwill of those who fundraised or paid for each one. Having people adopt individual trees and regularly watering them with the necessary 30 litres per week completely changes this narrative. 

A custom-built bird box houses the Pollinator Camera

Nurturing green shoots

Alongside the chatty trees, the Intelligent Garden featured a Raspberry Pi 5–based AI Pollinator Camera Trap on the roof of the main Chelsea Flower Show pavilion, which was covered in a meadow of wild grass. The bespoke bird box by Sebastian Cox is a thing of beauty in its own right. 

“This camera trap aims to replicate a Royal Horticultural Society initiative called FIT (Flower Insect Timed) counting, which aims to monitor the number and type of pollinating insects visiting a quadrant of garden space and landing on flowers so that the garden owner can assess whether they are attracting enough butterflies and bees,” says Intelligent Garden designer Pete.

The Pollinator Camera detects and logs the number of bees arriving in the flower garden transect

“By supplementing good horticulture and science with the power of AI, we hope to improve the chances of urban tree survival and to help maintain the existence of our much-needed urban green spaces.”

The Pollinator Camera recognises pollinators and keeps a running total, over a 24-hour period, of the number of insect detections in the past 30 minutes, along with thumbnail images of each insect or butterfly identified. “It can detect flowers, bees, and butterflies currently, and when insects appear in frame, we take still images as well as videos, both with bounding boxes to show the objects for use in the application,” Pete explains. The YOLO object detection setup currently records each instance, with tags added via Label Studio. Pete tried using AI to label them, but it was hit or miss, so he ended up doing this part manually. 

Installed on the pavilion roof, the camera automatically logs insect visitors and sends data back to Avanade’s server via a fixed Power-over-Ethernet connection

The next version of the Pollinator Camera software will identify individual bee species and insect types and record how many of each visit the garden. The live data means the thumbnails already being logged can be added to the dataset (or zoo) to train the next iteration of the software and improve its accuracy. “Of course, if you start training on an image that’s got a box around it, it’s going to be very quick to identify boxes, not bees. It’s the same for the video as well.” To get around this, Pete created “a little piping script” that captures one-second interval screenshots from the video. Avanade could use each of these “because of course, the position of the bee would change, and it would rotate, and it seemed important to train it on all of those”.

Avanade has provided detailed build instructions here, and the company urges makers and nature fans to build their own versions and share their findings. There’s even talk of a competition for the best young citizen science projects based on the one at the Chelsea Flower Show! 

The post Avanade Intelligent Garden at the Chelsea Flower Show appeared first on Raspberry Pi.

from News - Raspberry Pi https://ift.tt/yb4CvNU