Tag Archives: deep learning

xxx S. Kodors, M. Sondors, I. Apeinans, I. Zarembo, G. Lacis, E. Rubauskis and K. Karklina
Importance of mosaic augmentation for agricultural image dataset
Abstract |

Full text PDF (933 KB)

Importance of mosaic augmentation for agricultural image dataset

S. Kodors¹*, M. Sondors¹, I. Apeinans¹, I. Zarembo¹, G. Lacis², E. Rubauskis² and K. Karklina²

¹Rezekne Academy of Technologies, Faculty of Engineering, Institute of Engineering, Atbrivosanas Str. 115, LV-4601 Rezekne, Latvia
²Institute of Horticulture (LatHort), Graudu Str. 1, LV-3701 Cerini, Krimunu pagasts, Dobeles novads, Latvia
*Correspondence: sergejs.kodors@rta.lv

Abstract:

The yield estimation using artificial intelligence is based on object detection algorithms. Firstly, the object detection algorithms identify the number of fruits on images, then tree fruit load is predicted using regression algorithms. YOLO is a popular convolution neural network architecture for object detection tasks. It is sufficiently well studied for fruit yield estimation. However, the experiments are traditionally restricted to only one specific fruit category and growing season. This is a big shortcoming for the smart solutions like agro-drones, which must automatically complete yield monitoring of the most popular fruit species in commercial orchards. Therefore, the modern studies related to yield estimation increasingly raise attention to multi-stage, multi-state and multi-specie detection tasks. The multi-stage datasets can be described as a collection of multiple sub-datasets, e.g. flowers, fruitlets and fruits. The multi-state dataset can contain classes like mature, immature or damaged fruits. Meanwhile, the multi-specie dataset contains images with representatives of multiple cultures. However, if classic object-detection tasks like urban or indoor object detection have multiple classes presented in one image, then yield estimation datasets usually have images with only one class presented on them. Therefore, an image shuffle or mosaic augmentation are the intuitive training strategies of YOLO for object detection working with a collection of multiple single class datasets. We applied the YOLOv5m model to test both strategies, which were verified on three datasets: apple fruits (MinneApple), pear fruits (Pear640) and pear fruitlets (PFruitlet640). Our experiment showed that mosaic augmentation improves mAP@0.5:0.95 better than simple image shuffle. The mean difference between both strategies is equal to 0.0438.

Key words:

augmentation, deep learning, object detection, precision horticulture, yield estimation

1117–1133 K. Rajashree, K.V. Prema, G. Rajath and S. Angad
Prediction of fruit rot disease incidence in Arecanut based on weather parameters
Abstract |

Full text PDF (1 MB)

Prediction of fruit rot disease incidence in Arecanut based on weather parameters

K. Rajashree¹*, K.V. Prema², G. Rajath³ and S. Angad¹

¹Manipal Academy of Higher Education, Manipal Institute of Technology, Department of Computer Science and Engineering, Madhav Nagar, Manipal, IN576104 Karnataka India
²Manipal Academy of Higher Education, Manipal Institute of Technology, Department of Computer Science and Engineering, BSF Campus, Yelahanka, Bangalore, IN560064 Karnataka, India
³PES University, PES Institute of Technology, Department of Electronics and Communication, Dwaraka Nagar, Banashankari, Bangalore, IN560085 Karnataka, India
*Correspondence: raji.krish@manipal.edu

Abstract:

The occurrence of pests and diseases in arecanut crops has always been an important factor affecting the total production of arecanut. Arecanut is always dependent on environmental factors during its growth. Thus monitoring and early prediction of the occurrence of the disease would be very helpful for prevention and therefore more crop production. Here, we propose artificial intelligence-based deep learning models for fruit rot disease prediction. Historical data on fruit rot incidence in representative areas of arecanut production in Udupi along with historical weather data are the parameters used to develop region-specific models for the Udupi district. The fruit rot disease incidence score value is predicted using recurrent neural network variants (i.e., Vanilla LSTM, Vanilla GRU, stacked LSTM, and Bidirectional LSTM) for the first time. The predictive performance of the proposed models is evaluated by mean square error (MSE) along with the 5-fold cross-validation technique. Further, compared to other deep learning and machine learning models, the Vanilla LSTM model gives 1.5 MSE, while the Vanilla GRU model gives 1.3 MSE making it the best prediction model for arecanut fruit rot disease.

Key words:

crop disease, cross-validation, deep learning, fruit rot disease, prediction

507–519 S. Kodors, G. Lacis, O. Sokolova2,V. Zhukovs, I. Apeinans and T. Bartulsons
Apple scab detection using CNN and Transfer Learning
Abstract |

Full text PDF (2 MB)

Apple scab detection using CNN and Transfer Learning

S. Kodors¹*, G. Lacis², O. Sokolova2,V. Zhukovs¹, I. Apeinans¹ and T. Bartulsons²

¹Rezekne Academy of Technologies, Faculty of Engineering, Institute of Engineering, Atbrivoshanas Str. 115, LV-4601 Rezekne, Latvia
²Institute of Horticulture, Graudu Str. 1, LV-3701 Ceriņi, Krimūnu pagasts, Dobeles novads, Latvia
*Correspondence: sergejs.kodors@rta.lv

Abstract:

The goal of smart and precise horticulture is to increase yield and product quality by simultaneous reduction of pesticide application, thereby promoting the improvement of food security. The scope of this research is apple scab detection in the early stage of development using mobile phones and artificial intelligence based on convolutional neural network (CNN) applications. The research considers data acquisition and CNN training. Two datasets were collected – with images of scab infected fruits and leaves of an apple tree. However, data acquisition is a time-consuming process and scab appearance has a probability factor. Therefore, transfer learning is an appropriate training methodology. The goal of this research was to select the most suitable dataset for transfer learning for the apple scab detection domain and to evaluate the transfer learning impact comparing it with learning from scratch. The statistical analysis confirmed the positive effect of transfer learning on CNN performance with significance level 0.05.

Key words:

agriculture., artificial intelligence, deep learning, fungus, machine learning, Malus, pathogen, precise horticulture, Venturia