CropDeep: The Crop Vision Dataset for Deep-Learning-Based Classification and Detection in Precision Agriculture.

Yang-Yang Zheng,Jian-Lei Kong,Min Zuo,Xiao-Yi Wang,Xue-Bo Jin

doi:10.3390/s19051058

Yang-Yang Zheng, Jian-Lei Kong + Show 3 more

Open Access

https://doi.org/10.3390/s19051058

Copy DOI

Abstract

Intelligence has been considered as the major challenge in promoting economic potential and production efficiency of precision agriculture. In order to apply advanced deep-learning technology to complete various agricultural tasks in online and offline ways, a large number of crop vision datasets with domain-specific annotation are urgently needed. To encourage further progress in challenging realistic agricultural conditions, we present the CropDeep species classification and detection dataset, consisting of 31,147 images with over 49,000 annotated instances from 31 different classes. In contrast to existing vision datasets, images were collected with different cameras and equipment in greenhouses, captured in a wide variety of situations. It features visually similar species and periodic changes with more representative annotations, which have supported a stronger benchmark for deep-learning-based classification and detection. To further verify the application prospect, we provide extensive baseline experiments using state-of-the-art deep-learning classification and detection models. Results show that current deep-learning-based methods achieve well performance in classification accuracy over 99%. While current deep-learning methods achieve only 92% detection accuracy, illustrating the difficulty of the dataset and improvement room of state-of-the-art deep-learning models when applied to crops production and management. Specifically, we suggest that the YOLOv3 network has good potential application in agricultural detection tasks.

Highlights

Modern agriculture seeks to manage crops in controlled environments such as greenhouses, that are able to improve the production of plants or duplicate the environmental conditions of specific geographical areas to obtain imported products locally
We present a domain-specific vision dataset of crops, namely CropDeep, in the field of precision agriculture
The novelty of CropDeep is that it aims at providing the data benchmark to constructing deep-learning-based classification and detection models according to realistic characteristics of agricultural tasks in greenhouses

Summary

Introduction

Modern agriculture seeks to manage crops in controlled environments such as greenhouses, that are able to improve the production of plants or duplicate the environmental conditions of specific geographical areas to obtain imported products locally. It is possible to obtain highly accurate status of crops and form reasonable decisions to manage irrigation, change climate factors, or enrich the soil nutrition in agricultural scenes, which optimize automation of precise management and improve production of crops while potentially reducing environmental impacts [2]. With the sensor data obtained and transmitted by IoT, they use smartphones to remotely monitor their crops and equipment, understand the whole management status accurately with statistical analysis, and instruct the robots to carry out agricultural tasks. The greenhouses are taking advantage the integration of different technologies with efficient human intervention, the current level of artificial intelligence (AI) in agricultural machines and systems is far from achieving automated operations and management requiring minimum supervision to optimize production by accounting for variability and uncertainties within precision agriculture (PA) [4]

Methods

Results

Conclusion