| Image-based Discrimination of Ultrasound-Assisted Frozen Meat Using Inception-v3 |
| Paper ID : 1381-NICAME1402 |
| Authors: |
|
Mahmoud Soltani Firouz *1, Hamed Sardari2, Soleiman Hosseinpour1, Pouya Bohlol1 1Department of Agricultural Machinery Engineering, Faculty of Agriculture, University of Tehran, Karaj, Iran 2Faculty of Agriculture, University of Tehran |
| Abstract: |
| Categorizing images based on ultrasound power levels can be difficult because of the nuanced differences in image quality. Ultrasound technology utilizes high-frequency sound waves to generate cavitation bubbles within food structures, and the characteristics of these bubbles can vary depending on the power levels employed. The objective of this research is to investigate the possible applications of deep learning algorithms in categorizing sonicated meat samples according to their quality attributes, including cooking loss, thawing loss, cutting force, color, and texture. The findings could offer significant insights into the utilization of deep learning algorithms for analyzing meat quality data and their potential in creating intelligent meat processing systems. Through training a CNN model on a dataset of images categorized by ultrasound power levels, the model can identify and comprehend patterns and features related to specific power levels. The model achieves an accuracy rate of 97.75% for identifying thawed samples and 97.68% for cooked ones. The model's impressive performance in classifying the images highlights its capacity to accurately distinguish subtle variations in image quality and clarity. |
| Keywords: |
| Deep Learning; Ultrasound; GoogLeNet; Inception V3 |
| Status : Paper Accepted (Oral Presentation) |
