Real-world plant species identification based on deep convolutional neural networks and visual attention

Abstract

This paper investigates the issue of real-world identification to fulfill better species protection. We focus on plant species identification as it is a classic and hot issue. In tradition plant species identification the samples are scanned specimen and the background is simple. However, real-world species recognition is more challenging. We first systematically investigate what is realistic species recognition and the difference from tradition plant species recognition. To deal with the challenging task, an interdisciplinary collaboration is presented based on the latest advances in computer science and technology. We propose a novel framework and an effective data augmentation method for deep learning in this paper. We first crop the image in terms with visual attention before general recognition. Besides, we apply it as a data augmentation method. We call the novel data augmentation approach attention cropping (AC). Deep convolutional neural networks are trained to predict species from a large amount of data. Extensive experiments on traditional dataset and specific dataset for real-world recognition are conducted to evaluate the performance of our approach. Ex- * Corresponding author.periments first demonstrate that our approach achieves state-of-the-art results on different types of datasets. Besides, we also evaluate the performance of data augmentation method AC. Results show that AC provides superior performance. Compared with the precision of methods without AC, the results with AC achieve substantial improvement.

Central Attention and a Dual Path Convolutional Neural Network in Real-World Tree Species Recognition

“…Other nonsalient parts are ignored or neglected. Therefore, we are not even aware of the redundancy during the first judgement [19].…”

“…Xiao et al proposed a deep learning framework with attention cropping [19]. The input images were cropped in terms of visual attention before recognized.…”

Multiscale Convolutional Neural Networks with Attention for Plant Species Recognition

“…The pooling operation is a form of nonlinear downsampling, which can reduce the size of the feature maps extracted from the convolutional layers to achieve spatial invariance. The operation leads to faster convergence and improves the generalization performance [ 8 , 9 ]. When the feature map x j l is passed to the pooling layer, the pooling operation is applied to the feature map x j l , which produces a pooled feature map x j l +1 as the output.…”

“…e feature extraction of plant leaf image is a crucial step of a plant recognition method. ere are many plant species recognition algorithms [1][2][3][4][5], which can be divided into two main types of feature representation methods for describing leaf images, i.e., hand-crafted features [6,7] and deep learning features [8,9]. In fact, the hand-crafted features are mainly dependent on the ability of computer vision experts to encode the morphological characters of the leaves [2][3][4][5][6].…”

Data Augmentation using Adversarial Networks for Tea Diseases Detection

“…Works on [2] and [4] use the original image in the dataset as input for the experiment. Attention Cropping (AC) used in [5], transfer learning and deep feature extraction are used in [6], rotation transformation technique and mirror symmetry are used in [7], while [8] uses unsupervised feature learning. Only [8] uses DCGAN.…”