What question did this study set out to answer?

To create a large image dataset for weed species and establish baseline classification performance using deep learning models.

February 14, 2019Scientific ReportsOpen Access

DeepWeeds: A Multiclass Weed Species Image Dataset for Deep Learning

View Full Paper

Ask AI

Bookmark

Authors

Randomized trial assesses real-time performance of deep learning models for weed species classification, suggesting viability for robotic control.

To create a large image dataset for weed species and establish baseline classification performance using deep learning models.
Developed the DeepWeeds dataset with 17,509 images of eight weed species from northern Australia.
Evaluated classification performance using Inception-v3 and ResNet-50 models on the dataset.
Measured inference time for real-time classification using the ResNet-50 architecture.
Inception-v3 achieved an average classification accuracy of 95.1%.
ResNet-50 achieved an average classification accuracy of 95.7%.
Average inference time for ResNet-50 was 53.4 ms per image.

Olsen et al. (2019) studied this question.

View Full Paper

Ask AI

Bookmark

Randomized trial assesses real-time performance of deep learning models for weed species classification, suggesting viability for robotic control.