Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Efficient Microscopic Image Instance Segmentation for Food Crystal Quality Control

Xiaoyu Ji, Jan P Allebach, Ali Shakouri, Fengqing Zhu

TL;DR

Experimental results show that the predicted crystal counting accuracy of the proposed efficient instance segmentation method is comparable with existing segmentation methods, while being five times faster.

Abstract

This paper is directed towards the food crystal quality control area for manufacturing, focusing on efficiently predicting food crystal counts and size distributions. Previously, manufacturers used the manual counting method on microscopic images of food liquid products, which requires substantial human effort and suffers from inconsistency issues. Food crystal segmentation is a challenging problem due to the diverse shapes of crystals and their surrounding hard mimics. To address this challenge, we propose an efficient instance segmentation method based on object detection. Experimental results show that the predicted crystal counting accuracy of our method is comparable with existing segmentation methods, while being five times faster. Based on our experiments, we also define objective criteria for separating hard mimics and food crystals, which could benefit manual annotation tasks on similar dataset.

Efficient Microscopic Image Instance Segmentation for Food Crystal Quality Control

TL;DR

Experimental results show that the predicted crystal counting accuracy of the proposed efficient instance segmentation method is comparable with existing segmentation methods, while being five times faster.

Abstract

This paper is directed towards the food crystal quality control area for manufacturing, focusing on efficiently predicting food crystal counts and size distributions. Previously, manufacturers used the manual counting method on microscopic images of food liquid products, which requires substantial human effort and suffers from inconsistency issues. Food crystal segmentation is a challenging problem due to the diverse shapes of crystals and their surrounding hard mimics. To address this challenge, we propose an efficient instance segmentation method based on object detection. Experimental results show that the predicted crystal counting accuracy of our method is comparable with existing segmentation methods, while being five times faster. Based on our experiments, we also define objective criteria for separating hard mimics and food crystals, which could benefit manual annotation tasks on similar dataset.
Paper Structure (17 sections, 1 equation, 5 figures, 3 tables)

This paper contains 17 sections, 1 equation, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Instance segmentation
  4. Food crystal image analysis
  5. Method
  6. Object detection model
  7. Post processing
  8. Dataset and Manual Annotation Criterion
  9. Seed count food crystal dataset
  10. Manual annotation criteria
  11. Experiments
  12. Experimental Setting
  13. Quantitative Results
  14. Processing Time Comparison
  15. Qualitative Comparison
  16. ...and 2 more sections

Figures (5)

  • Figure 1: Hard mimics examples. The objects in yellow boxes are hard mimics. Crystals and air bubbles are in blue and green boxes, respectively.
  • Figure 2: Overview of our proposed method. "Bbox" represents bounding box, "cls" represents class. The images from left to right are an example input image and the results of each step.
  • Figure 3: Example food crystal images captured from the microscope. (a) original image (lightness adjusted), (b) manual annotation result. Crystals are marked with blue, air bubbles are marked with yellow, while hard mimics are marked with red.
  • Figure 4: Visualization comparisons of different segmentation methods on food crystal dataset. From left to right are the input images, the ground truth label images, results of Mask RCNN he2017, Mask-scoring RCNN huang2019mask, Stardist schmidt2018cell, Yolov8 segmentaton yolov8 and our method.
  • Figure 5: Normalized confusion matrix for the object detection model.