Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

RoWeeder: Unsupervised Weed Mapping through Crop-Row Detection

Pasquale De Marinis, Gennaro Vessio, Giovanna Castellano

TL;DR

RoWeeder is an innovative framework for unsupervised weed mapping that combines crop-row detection with a noise-resilient deep learning model that trains a lightweight deep learning model capable of distinguishing between crops and weeds, even in the presence of noisy data.

Abstract

Precision agriculture relies heavily on effective weed management to ensure robust crop yields. This study presents RoWeeder, an innovative framework for unsupervised weed mapping that combines crop-row detection with a noise-resilient deep learning model. By leveraging crop-row information to create a pseudo-ground truth, our method trains a lightweight deep learning model capable of distinguishing between crops and weeds, even in the presence of noisy data. Evaluated on the WeedMap dataset, RoWeeder achieves an F1 score of 75.3, outperforming several baselines. Comprehensive ablation studies further validated the model's performance. By integrating RoWeeder with drone technology, farmers can conduct real-time aerial surveys, enabling precise weed management across large fields. The code is available at: \url{https://github.com/pasqualedem/RoWeeder}.

RoWeeder: Unsupervised Weed Mapping through Crop-Row Detection

TL;DR

RoWeeder is an innovative framework for unsupervised weed mapping that combines crop-row detection with a noise-resilient deep learning model that trains a lightweight deep learning model capable of distinguishing between crops and weeds, even in the presence of noisy data.

Abstract

Precision agriculture relies heavily on effective weed management to ensure robust crop yields. This study presents RoWeeder, an innovative framework for unsupervised weed mapping that combines crop-row detection with a noise-resilient deep learning model. By leveraging crop-row information to create a pseudo-ground truth, our method trains a lightweight deep learning model capable of distinguishing between crops and weeds, even in the presence of noisy data. Evaluated on the WeedMap dataset, RoWeeder achieves an F1 score of 75.3, outperforming several baselines. Comprehensive ablation studies further validated the model's performance. By integrating RoWeeder with drone technology, farmers can conduct real-time aerial surveys, enabling precise weed management across large fields. The code is available at: \url{https://github.com/pasqualedem/RoWeeder}.
Paper Structure (10 sections, 5 equations, 4 figures, 5 tables)

This paper contains 10 sections, 5 equations, 4 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. RoWeeder
  4. Experimental Evaluation
  5. Dataset and Validation
  6. Setting
  7. Results
  8. Ablation Study
  9. Conclusion
  10. Acknowledgement

Figures (4)

  • Figure 1: Overview of the proposed framework, illustrating the training process of the semantic segmentation model. In the top branch, the image is fed into a plant detection algorithm, and the mask produced is used to detect crop rows. Every plant on a crop row is classified as a crop; otherwise, it is classified as a weed. This pseudo-ground truth is used to train a deep learning model.
  • Figure 2: Visual representation of the crop lines detected by the Hough Transform. From left to right: the original input image, the ground truth with crops in green and weeds in red, and the generated pseudo-ground truth with detected crop rows in purple.
  • Figure 3: Overview of the two decoders. On top is the pyramid decoder with concatenation as a fusion method. On the bottom is the flat decoder with sum as the fusion method. The input feature maps are ordered from the most shallow to the deeper ones. "Spatial Conv" is a $3\times3$ convolution, while "Point Conv" is a $1 \times 1$ convolution.
  • Figure 4: Pseudo-ground truth is generated with the Hough+CC method and prediction by RoWeeder. The background is denoted in black, crops in green, and weeds in red.