Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

M18K: A Comprehensive RGB-D Dataset and Benchmark for Mushroom Detection and Instance Segmentation

Abdollah Zakeri, Mulham Fawakherji, Jiming Kang, Bikram Koirala, Venkatesh Balan, Weihang Zhu, Driss Benhaddou, Fatima A. Merchant

TL;DR

This paper tackles the lack of public datasets for edible mushroom detection and segmentation in smart agriculture by introducing M18K, a comprehensive RGB-D dataset comprising 423 image pairs with 18,172 annotated mushroom instances across white button and baby bella varieties captured by an Intel RealSense D405. It details a SAM-based labeling workflow followed by manual refinement, and provides depth maps, 3D point clouds, and precise instance-level masks to support detection and segmentation benchmarking. The authors evaluate 11 models on 1280×720 RGB data, revealing that depth information offers limited gains due to reinitialization costs, with Mask R-CNN (ResNet-50) and RT-DETR-L delivering strong performance at different tasks. By releasing the dataset, code, and trained models publicly, the work enables reproducible benchmarking and accelerates development of automated harvesting, growth monitoring, and disease detection in edible mushroom farming.

Abstract

Automating agricultural processes holds significant promise for enhancing efficiency and sustainability in various farming practices. This paper contributes to the automation of agricultural processes by providing a dedicated mushroom detection dataset related to automated harvesting, growth monitoring, and quality control of the button mushroom produced using Agaricus Bisporus fungus. With over 18,000 mushroom instances in 423 RGB-D image pairs taken with an Intel RealSense D405 camera, it fills the gap in mushroom-specific datasets and serves as a benchmark for detection and instance segmentation algorithms in smart mushroom agriculture. The dataset, featuring realistic growth environment scenarios with comprehensive annotations, is assessed using advanced detection and instance segmentation algorithms. The paper details the dataset's characteristics, evaluates algorithmic performance, and for broader applicability, we have made all resources publicly available including images, codes, and trained models via our GitHub repository https://github.com/abdollahzakeri/m18k

M18K: A Comprehensive RGB-D Dataset and Benchmark for Mushroom Detection and Instance Segmentation

TL;DR

This paper tackles the lack of public datasets for edible mushroom detection and segmentation in smart agriculture by introducing M18K, a comprehensive RGB-D dataset comprising 423 image pairs with 18,172 annotated mushroom instances across white button and baby bella varieties captured by an Intel RealSense D405. It details a SAM-based labeling workflow followed by manual refinement, and provides depth maps, 3D point clouds, and precise instance-level masks to support detection and segmentation benchmarking. The authors evaluate 11 models on 1280×720 RGB data, revealing that depth information offers limited gains due to reinitialization costs, with Mask R-CNN (ResNet-50) and RT-DETR-L delivering strong performance at different tasks. By releasing the dataset, code, and trained models publicly, the work enables reproducible benchmarking and accelerates development of automated harvesting, growth monitoring, and disease detection in edible mushroom farming.

Abstract

Automating agricultural processes holds significant promise for enhancing efficiency and sustainability in various farming practices. This paper contributes to the automation of agricultural processes by providing a dedicated mushroom detection dataset related to automated harvesting, growth monitoring, and quality control of the button mushroom produced using Agaricus Bisporus fungus. With over 18,000 mushroom instances in 423 RGB-D image pairs taken with an Intel RealSense D405 camera, it fills the gap in mushroom-specific datasets and serves as a benchmark for detection and instance segmentation algorithms in smart mushroom agriculture. The dataset, featuring realistic growth environment scenarios with comprehensive annotations, is assessed using advanced detection and instance segmentation algorithms. The paper details the dataset's characteristics, evaluates algorithmic performance, and for broader applicability, we have made all resources publicly available including images, codes, and trained models via our GitHub repository https://github.com/abdollahzakeri/m18k
Paper Structure (9 sections, 6 figures, 2 tables)

This paper contains 9 sections, 6 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Mushroom Detection and Localization
  4. Mushroom Growth Monitoring And Quality Assessment
  5. Mushroom Disease Detection
  6. Labeling Process
  7. Data Description
  8. Benchmarking and Results
  9. Conclusion

Figures (6)

  • Figure 1: Sample of dataset images, ground truth instance segmentation masks, and depth images for Baby Bella (BB) and White Button (WB) mushroom images
  • Figure 2: RGB-D Point Cloud Sample
  • Figure 3: Scatter plots of instances mask area vs bounding box diagonal length for (A) baby bella mushrooms and (B) white button mushrooms.
  • Figure 4: Histograms of Mask Areas and Bounding Box Diagonal Lengths; Histograms (A) and (B) show pixel mask areas of Baby Bella (BB) and White Button (WB) mushrooms accordingly. Histograms (C) and (D) show bounding box diagonal lengths for BB and WB mushroom instances accordingly.
  • Figure 5: Color distribution of mushroom instances in various color spaces; Judging only by the color distributions, LAB and HSV color spaces are more suitable than RGB for classifying WB and BB instances. WB mushrooms are shown in orange and BB mushrooms are shown in blue.
  • ...and 1 more figures