Improved Ice Cream Detection Algorithm Based on YOLOv5

Jingang Ma, Yong JiangICML 2022 · DOI

Abstract

Manufacturers make production decisions by analyzing the sales of different types of ice cream at each retail store from user purchase videos. The color difference in the appearance of ice cream in the video can cause missed detection of objects, therefore, the instance normalization and batch normalization modules is introduced into YOLOV5-x; and then, to reduce the semantic feature loss caused by up-sampling, the flow alignment module is used to improve detection performance. The experimental results indicate that, compared with the original YOLOv5-x, our method increases by 1.2% on [email protected] and increases by 0.7% on [email protected].
BibTeX
@inproceedings{ma2022improved,
  author = {Jingang Ma and Yong Jiang},
  title = {{Improved Ice Cream Detection Algorithm Based on YOLOv5}},
  booktitle = {International Conference on Machine Learning},
  year = {2022},
  doi = {10.1109/mlise57402.2022.00040},
}

Loading related content...

Figure