Classification of plant seedlings using deep convolutional neural network architectures

Abstract

Weed management has a vital role in applications of agriculture domain. One of the key tasks is to identify the weeds after few days of plant germination which helps the farmers to perform early-stage weed management to reduce the contrary impacts on crop growth. Thus, we aim to classify the seedlings of crop and weed species. In this work, we propose a plant seedlings classification using the benchmark plant seedlings dataset. The dataset contains the images of 12 different species where three belongs to plant species and the other nine belongs to weed species. We implement the classification framework using three different deep convolutional neural network architectures, namely ResNet50V2, MobileNetV2 and EfficientNetB0. We train the models using transfer learning and compare the performance of each model on a test dataset of 833 images. We compare the three models and demonstrate that the EfficientNetB0 performs better with an average F1-Score of 96.26% and an accuracy of 96.52%.

Deep Convolutional Neural Network Architecture for Plant Seedling Classification

“…For the Efficient Net B2 model, classes black grass, common wheat, fat-hen, loose silky bent, and sugar beet have F1-score of 0.85, 0.88, 0.98, 0.94, and 0.97 while the remaining plant species are correctly classified. Figures 7 and 8 The proposed Efficient Net B2 and Efficient Net B4 methods were compared with existing methods [16][17][18][19] for the plant seedlings dataset and the results are shown in Table III. We can see that the proposed Efficient Net B4 method outperforms the latest competitive approaches in terms of accuracy and F1-score.…”

“…The dataset was obtained from Computer Vision and Biosystems Signal Processing group, Aarhus University, an open-source website which contains plant seedling images. The dataset contains 5541 plant seedling images, namely 833 test images, 833 validation images, and 3875 training images were acquired, each representing a different stage of growth for all the 12 considered species [1,19].…”

“…The motivation of our research work was to investigate the Efficient Net [19,27] architecture because this model outperforms other CNN models in terms of accuracy and efficiency. The Efficient Net model has received little attention when considering datasets of plant seedlings.…”

“…depth, width, and resolution) of the network by having a balanced output for the above parameters. A compound scaling technique is suggested in [19] that consistently scales width, depth, and resolution according to a set of guiding principles. They suggest the following formula:…”

“…Weeds and crops were classified using Residual Network 101 in [18], obtaining an overall accuracy rate of 98.47% on the validation set and 96.04% on the test set. CNN frameworks for plant seedlings classification by using ResNet50V2, MobileNetV2, and Efficient Net B0 architectures were implemented in [19]. The Efficient Net B0 method provided an average F1-Score of 96.26% and an accuracy of 96.52%.…”