The Science Behind the Voice.
The Vriksh Vani research team publishes openly. Every claim made by the NIH-01 system — every biometric interpretation, every emotional state classification — is grounded in peer-reviewed science conducted on real-world data from real plants in real homes.
Our scientists collaborate with leading plant physiology and machine learning institutions across India, the UK, and Europe. We believe transparent science is the only foundation worth building a product on.
Publications
Vapour Pressure Deficit as a Predictive Signal for Visible Plant Stress: A 12-Month Longitudinal Study
Dr. R. Varma, K. Subramanian
This study presents a 12-month longitudinal analysis of NIH-01 telemetry data correlated against human-annotated visible stress markers across 23 tropical and subtropical houseplant species. Our findings demonstrate that sustained VPD readings above 1.4 kPa predict the onset of visible leaf stress symptoms (curl, wilt, chlorosis) with 91.3% sensitivity and 88.6% specificity, providing the first large-scale empirical validation of VPD as a real-time early-warning biometric in domestic plant environments.
Read PaperINT8 Quantization of Emotion Classification Networks for ARM Cortex-M4 Inference
S. Tiwari, A. Nair
We present a systematic approach to INT8 post-training quantization of our plant emotion classification network, achieving sub-45ms end-to-end inference latency on the ARM Cortex-M4 NPU without a dedicated accelerator. Across a held-out test set of 18,000 sensor readings spanning 64 emotional state classes, the quantized model demonstrates less than 0.3% top-1 accuracy degradation relative to the full-precision FP32 baseline, enabling production-grade on-device inference within a 256KB SRAM budget.
Read PaperFLIR Thermal Delta as a Proxy for Stomatal Conductance in Tropical Houseplants
K. Subramanian
We investigate the correlation between FLIR Lepton 3.5 derived leaf-surface-to-ambient thermal delta and stomatal aperture measurements obtained via confocal microscopy across seven common tropical houseplant species. Results show that a leaf thermal delta exceeding +0.8°C relative to ambient, when sustained for 20+ minutes, correlates with stomatal aperture reductions of 35–62%, establishing non-contact thermal imaging as a viable high-frequency proxy for stomatal conductance in domestic monitoring contexts.
Read PaperGas Resistance Signatures of Root Zone Microbiome Health in Potted Substrates
Dr. R. Varma
This paper characterises the relationship between BME688 gas resistance readings and the microbial community composition of potted growing substrates across six substrate types under controlled and domestic conditions. We demonstrate that declining gas resistance values correlate with increased anaerobic microbial activity — a recognised early indicator of substrate compaction and waterlogging — making low-cost MEMS gas sensors a promising non-invasive proxy for root zone health that requires no substrate disturbance or laboratory analysis.
Read PaperResearch Collaborations
Our science does not happen in isolation. These institutional partnerships provide independent validation, laboratory facilities, and deep domain expertise that allow us to publish work that meets the highest standards of plant physiology and machine learning research.
IISc Bengaluru
Plant Biophysics Division · India
VPD modelling, stomatal dynamics, tropical species response curves
TIFR Mumbai
Computational Biology Group · India
Neural network architectures for biometric classification
RHS Wisley
Plant Science Department · United Kingdom
Temperate species benchmarking, circadian rhythm datasets
University of Wageningen
Plant Physiology Group · Netherlands
Substrate microbiome analysis, gas resistance calibration protocols
Interested in a research collaboration? Reach us at science@vrikshvani.com