Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Learning for Precision Agriculture: Post-Spraying Evaluation and Deposition Estimation

Harry Rogers, Tahmina Zebin, Grzegorz Cielniak, Beatriz De La Iglesia, Ben Magri

TL;DR

This work tackles automated post-spraying evaluation in precision agriculture without traditional tracers or WSPs. It introduces an XAI CV pipeline that performs semantic segmentation of sprayed targets across seven classes and computes class-wise spray deposition through a domain-specific WSDE task, aided by inference-only feature fusion for better interpretability. The results show that DeepLabV3 with EfficientNet-B0 achieves top segmentation metrics with ADD fusion, while FCN with ADD fusion and Affinity Propagation provides the strongest WSDE performance, achieving a deposition error of $156.8$ $\mu$L and a mean hit rate of $38.3\%$; the dataset is publicly released to support reproducibility. Overall, the method enables automated, interpretable post-spray evaluation and deposition estimation, potentially reducing reliance on manual methods and enabling scalable precision spraying quality assurance.

Abstract

Precision spraying evaluation requires automation primarily in post-spraying imagery. In this paper we propose an eXplainable Artificial Intelligence (XAI) computer vision pipeline to evaluate a precision spraying system post-spraying without the need for traditional agricultural methods. The developed system can semantically segment potential targets such as lettuce, chickweed, and meadowgrass and correctly identify if targets have been sprayed. Furthermore, this pipeline evaluates using a domain-specific Weakly Supervised Deposition Estimation task, allowing for class-specific quantification of spray deposit weights in μL. Estimation of coverage rates of spray deposition in a class-wise manner allows for further understanding of effectiveness of precision spraying systems. Our study evaluates different Class Activation Mapping techniques, namely AblationCAM and ScoreCAM, to determine which is more effective and interpretable for these tasks. In the pipeline, inference-only feature fusion is used to allow for further interpretability and to enable the automation of precision spraying evaluation post-spray. Our findings indicate that a Fully Convolutional Network with an EfficientNet-B0 backbone and inference-only feature fusion achieves an average absolute difference in deposition values of 156.8 μL across three classes in our test set. The dataset curated in this paper is publicly available at https://github.com/Harry-Rogers/PSIE

Deep Learning for Precision Agriculture: Post-Spraying Evaluation and Deposition Estimation

TL;DR

This work tackles automated post-spraying evaluation in precision agriculture without traditional tracers or WSPs. It introduces an XAI CV pipeline that performs semantic segmentation of sprayed targets across seven classes and computes class-wise spray deposition through a domain-specific WSDE task, aided by inference-only feature fusion for better interpretability. The results show that DeepLabV3 with EfficientNet-B0 achieves top segmentation metrics with ADD fusion, while FCN with ADD fusion and Affinity Propagation provides the strongest WSDE performance, achieving a deposition error of L and a mean hit rate of ; the dataset is publicly released to support reproducibility. Overall, the method enables automated, interpretable post-spray evaluation and deposition estimation, potentially reducing reliance on manual methods and enabling scalable precision spraying quality assurance.

Abstract

Precision spraying evaluation requires automation primarily in post-spraying imagery. In this paper we propose an eXplainable Artificial Intelligence (XAI) computer vision pipeline to evaluate a precision spraying system post-spraying without the need for traditional agricultural methods. The developed system can semantically segment potential targets such as lettuce, chickweed, and meadowgrass and correctly identify if targets have been sprayed. Furthermore, this pipeline evaluates using a domain-specific Weakly Supervised Deposition Estimation task, allowing for class-specific quantification of spray deposit weights in μL. Estimation of coverage rates of spray deposition in a class-wise manner allows for further understanding of effectiveness of precision spraying systems. Our study evaluates different Class Activation Mapping techniques, namely AblationCAM and ScoreCAM, to determine which is more effective and interpretable for these tasks. In the pipeline, inference-only feature fusion is used to allow for further interpretability and to enable the automation of precision spraying evaluation post-spray. Our findings indicate that a Fully Convolutional Network with an EfficientNet-B0 backbone and inference-only feature fusion achieves an average absolute difference in deposition values of 156.8 μL across three classes in our test set. The dataset curated in this paper is publicly available at https://github.com/Harry-Rogers/PSIE
Paper Structure (19 sections, 2 equations, 7 figures, 6 tables)

This paper contains 19 sections, 2 equations, 7 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Precision Spraying evaluation
  4. XAI methodologies
  5. Dataset
  6. XY spot sprayer
  7. Semi-automatic segmentation annotation
  8. Deposition Estimation
  9. XAI Pipeline
  10. Segmentation Architectures
  11. Inference-Only feature fusion
  12. Weakly Supervised Deposition Estimation
  13. Segmentation Metrics
  14. CAM Metrics
  15. Results
  16. ...and 4 more sections

Figures (7)

  • Figure 1: Example of a single spray deposit.
  • Figure 2: Semi automatic process for segmentation annotation using SAM.
  • Figure 3: RGB image \ref{['fig2a']} against ground truth semantic segmentation \ref{['fig2b']}.
  • Figure 4: Image labelled with keypoints for Weakly Supervised Deposition Estimation using the centre of spray actuations.
  • Figure 5: Prediction from Deep Learning model \ref{['pred']} with CAM for sprayed chickweed \ref{['cam']} with resulting island images after thresholding and combination in \ref{['islands']}.
  • ...and 2 more figures