PLNet: Light Recipe Design for Indoor Farming through Generative Deep Learning

Tanvi Verma

Abstract:

—Indoor farming has emerged as a promising solution
for year-round cultivation and efficient resource utilization in
crop production. Achieving optimal plant growth and quality in
indoor environments requires precise control of light conditions.
This study introduces PLNet, a generative deep learning method
for the designing the light recipes specifically tailored for indoor
farming.
Leveraging the power of deep neural networks, our approach
establishes complex connections between light spectra and plant
growth characteristics. Initially, a biomass estimator model is
trained using a diverse dataset encompassing different light
recipes and corresponding plant responses. Subsequently, a
generative model is trained using the estimator model as a foundation, enabling the generation of optimal light spectra to achieve
desired growth outcomes. This novel generative method offers an
efficient and effective approach to formulating light recipes for
indoor farming. By reducing the reliance on traditional trial-anderror methods, our method saves significant time and resources.
The presented generative deep learning method holds great
potential for advancing the design of light recipes in indoor
farming. Leveraging the capabilities of deep neural networks
facilitates more targeted and efficient optimization of light
conditions, resulting in improved crop yield and quality for a
variety of leafy green crops. The findings of this study contribute
to the ongoing efforts in enhancing productivity and sustainability
in indoor cultivation practices.