This paper studies the use of deep-learning models (AlexNet, VggNet, ResNet) pre-trained on object categories (ImageNet) in applied texture classification problems such as plant disease detection tasks. Research related to precision agriculture is of high relevance due to its potential economic impact on agricultural productivity and quality. Within this context, we propose a deep learning-based feature extraction method for the identification of plant species and the classification of plant leaf diseases. We focus on results relevant to real-time processing scenarios that can be easily transferred to manned/unmanned agricultural smart machinery (e.g. tractors, drones, robots, IoT smart sensor networks, etc.) by reconsidering the common processing pipeline. In our approach, texture features are extracted from different layers of pre-trained Convolutional Neural Network models and are later applied to a machine-learning classifier. For the experimental evaluation, we used publicly available datasets consisting of RGB textured images and datasets containing images of healthy and non-healthy plant leaves of different species. We compared our method to feature vectors derived from traditional handcrafted feature extraction descriptors computed for the same images and end-to-end deep-learning approaches. The proposed method proves to be significantly more efficient in terms of processing times and discriminative power, being able to surpass traditional and end-toend CNN-based methods and provide a solution also to the problem of the reduced datasets available for precision agriculture.
Abstract. Image filtering is one of the principal tools used in computer vision applications. Real systems store and manipulate high resolution images in compressed forms, therefore the implementation of the entire processing chain directly in the compressed domain became essential. This includes almost always linear filtering operations implemented by convolution. Linear image filtering implementation directly on the JPEG images is challenging for several reasons, including the complexity of transposing the pixel level convolution in the compressed domain, which may increase the processing time, despite avoiding the decompression. In this paper we propose a new computationally efficient solution for JPEG image filtering (as a spatial convolution between the input image and a given kernel) directly in the DCT based compressed domain. We propose that the convolution operation to be applied just on the periodical extensions of the DCT basis images, as an off-line processing, obtaining the filtered DCT basis images, which are used in data decompression. While this doesn't solve the near block boundaries filtering artefacts for large convolution kernels, for most practical cases, it provides good quality results at a very low computational complexity. These kind of implementations can run at real-time rates/ speeds and are suitable for developments of applications on digital cameras/ DSP/ FPGA.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.