Turning spectra into images improves plant trait retrieval with 2D-CNNs

2026-08-17

AI Summary

This study addresses the limitation of one-dimensional deep learning in capturing long-range dependencies within spectral data by proposing a 2D spectral reshaping strategy. Converting spectra into 2D images, this work integrates 2D-CNNs with masked autoencoders for multi-trait prediction. Results demonstrate that simple grid transformations outperform complex architectures, with model interpretability validated via Integrated Gradients. The direct reshaping method achieved an R² of 0.684, surpassing state-of-the-art performance by 0.097. Furthermore, self-supervised pretraining significantly exceeded 1D baselines, and identified key wavelength importance aligns with radiative transfer models. Collectively, these findings confirm that 2D reshaping effectively overcomes the constraints of traditional sequential processing in spectroscopic analysis.

Abstract

Hyperspectral reflectance spectroscopy enables non-destructive estimation of plant functional traits, yet current deep learning approaches process spectra as one-dimensional sequences, which limits how they capture long-range inter-band dependencies. We asked whether transforming 1D spectra into 2D image representations improves multi-trait prediction with convolutional neural networks (CNN). We compared nine transformations using EfficientNet-B0 on the GreenHyperSpectra dataset (7,897 labeled spectra, eight traits, 400-2450 nm), benchmarked against published 1D CNN results on the same split. Trained from scratch, the simplest transformation, a direct Reshape of the spectrum into a 2D grid, performed best ($R^2 = 0.684 \pm 0.001$) and improved on the state-of-the-art 1D baseline ($R^2 = 0.587$, $+0.097$). We then pretrained a 2D masked autoencoder (MAE-2D) on 139,000 unlabeled spectral images. Linear probing, which freezes the encoder and trains only a multilayer perceptron head, reached $R^2 = 0.646$ and exceeded every 1D self-supervised counterpart, including the fine-tuned MAE-1D ($R^2 = 0.641$). Under cross-dataset evaluation all models lost most of their accuracy and none beat the 1D baseline significantly. To identify which wavelengths drive each prediction, we applied Integrated Gradients and Grad-CAM and unfolded band importance back to the spectral axis. Protein ($r = 0.45$) and leaf water ($r = 0.33$) agreed with sensitivities simulated by the PROSAIL radiative-transfer model, while carotenoids ($r = 0.06$) and leaf area index ($r = -0.11$) did not, showing that the model reads established leaf chemistry for traits with sharp absorption features. The representational advantage of 2D spectral images, rather than architectural complexity or ImageNet pretraining, drives the gain over 1D approaches.

Problem

• Hyperspectral reflectance spectroscopy
• Plant functional traits
• Long-range inter-band dependencies
• 1D spectral processing
• Multi-trait prediction

Innovation

• 2D spectral representation
• masked autoencoder
• plant trait retrieval
• hyperspectral imaging
• explainable AI