Dr. Plant

• 

  Logo

Inspiration

Singapore has a bold plan call the ‘30 by 30’ goal, which is to build up the nation’s agri-food industry’s capability and capacity to produce 30% of the nutritional needs locally and sustainably by 2030.

To complement this goal, NParks also launched the ‘Gardening with Edibles’ programme, an initiative to encourage the local community in gardening.

What it does

Dr. Plant does 2 tasks.

Task 1 assess the plant's external conditions: the prototype is to record readings on external environment, such as the temperature, pH levels of soils and nitrogen/oxygen levels, and give an assessment on the plant's overall conditions based on these readings. Thereafter suggest to user on ways to improve the conditions.

Task 2 monitor the plant's health: the prototype is to monitor plant's leaves and to detect if the leaves show symptoms where the plant is not growing healthily, so that the prototype can suggest to user ways to treat the plans if it is unhealthy.

How we built it

We use the sensors attached to the raspberry pi to collect data such as the humidity, pH level of the soil and the rainfall and train a neural network to predict the probability scores for all possible plants, so that the gardener knows with all these conditions, what’s the best plant to grow. A separate pre-trained computer vision model Xception is used to detect the leaf of the plant and classify whether the plant has disease and what is the disease. Then the predictions from both model will be integrated into a JSON API and is being sent and display into the web app dashboard for monitoring.

Challenges we ran into

Incorporating the Raspberry Pi with the Web App for the prototyping

Accomplishments that we're proud of

• fully developed app

What we learned

Get more information of our plants.

What's next for Dr. Plant

For future plans, we aim to incorporate Machine Learning Operation, which is MLOps to keep track of the Machine Learning life cycle. In the event where there is data drift or concept drift, the deep learning models can be retrained constantly, such as including a new species of plants. Also, cloud-based technologies like AWS can be used to scale up our model to incorporate more species into our models. Lastly, we look at the business model where we can rent the equipment to the urban gardeners and we take a certain amount of commissions from their earnings for this scheme, that way we can ensure that our business is sustainable

Built With

• css
• javascript
• python
• rpi