Plant Biomass Estimation

Estimating plant biomass from pasture images using deep learning and multi-target regression to support animal forage assessment and ecological monitoring.

Back to home View GitHub
Image of plant biomass estimation model output

Project overview

A computer vision workflow that predicts multiple biomass targets from plant imagery using EfficientNet-based feature extraction.

  • Processed pasture image data for plant growth and biomass signals
  • Built a multi-target regression model for several biomass metrics
  • Focused on transfer learning and model robustness for real-world field data

Approach

Leveraged EfficientNet for high-quality visual features and trained a regression head to estimate plant biomass across multiple outputs.

Deep Learning Computer Vision EfficientNet Regression Transfer Learning

Results

Delivered accurate biomass predictions with a model well-suited for downstream pasture management and research workflows.

Final evaluation demonstrated strong correlation between predicted and actual biomass measurements while keeping inference efficient enough for practical deployment.