This paper presents a full pipeline to classify sample sets of corn kernels. The proposed approach follows a segmentation-classification scheme. The image segmentation is performed through a well known deep learning based approach, the Mask R-CNN architecture, while the classification is performed by means of a novel-lightweight network specially designed for this task—good corn kernel, defective corn kernel and impurity categories are considered. As a second contribution, a carefully annotated multitouching corn kernel dataset has been generated. This dataset has been used for training the segmentation and the classification modules. Quantitative evaluations have been performed and comparisons with other approaches provided showing improvements with the proposed pipeline.
This dataset contains 3 types of images:
1. Used for segmentation with 23 images (separated into 16 for training, 3 for testing, and 4 for validation). The images in this dataset have a resolution of 1280x1024.
2. Used for classification of 3 classes with 6600 images (separated into 3600 for training, 2100 for testing, and 900 for validation). The images in this dataset have a resolution of 224x224.
3. Used for classification of 2 classes with 3570 images (separated in 2310 for training, 360 for testing, and 900 for validation). The images in this dataset have a resolution of 224x224. (paper)
Download Dataset |
@inproceedings{corn2020ckcnn,
title={Deep Learning based Corn Kernel Classification},
author={Velesaca, Henry O. and Mira, Raúl and Suárez, Patricia L. and Larrea, Christian X. and Sappa, Angel D.},
booktitle={The 1st International Workshop on Agriculture-Vision: Challenges & Opportunities for Computer Vision in Agriculture},
pages={},
year={2020},
organization={}
}