We propose a deep learning approach to free-hand sketch recognition that achieves state-of-the-art performance, significantly surpassing that of humans. Our superior performance is a result of ...modelling and exploiting the unique characteristics of free-hand sketches, i.e., consisting of an ordered set of strokes but lacking visual cues such as colour and texture, being highly iconic and abstract, and exhibiting extremely large appearance variations due to different levels of abstraction and deformation. Specifically, our deep neural network, termed Sketch-a-Net has the following novel components: (i) we propose a network architecture designed for sketch rather than natural photo statistics. (ii) Two novel data augmentation strategies are developed which exploit the unique sketch-domain properties to modify and synthesise sketch training data at multiple abstraction levels. Based on this idea we are able to both significantly increase the volume and diversity of sketches for training, and address the challenge of varying levels of sketching detail commonplace in free-hand sketches. (iii) We explore different network ensemble fusion strategies, including a re-purposed joint Bayesian scheme, to further improve recognition performance. We show that state-of-the-art deep networks specifically engineered for photos of natural objects fail to perform well on sketch recognition, regardless whether they are trained using photos or sketches. Furthermore, through visualising the learned filters, we offer useful insights in to where the superior performance of our network comes from.
•Using manifold learning to transform original logical label space to Euclidean label space.•The similarity between samples is constrained by the similarity of corresponding numerical labels.•The ...final selection criterion integrates the influence of both the supervision information and local properties of the data.
In recent years, multi-label learning has been increasingly applied to various application areas. As an important pre-processing technique for multi-label learning, multi-label feature selection selects meaningful features to improve classification performance. In this paper, a feature selection method named manifold-based constraint Laplacian score (MCLS) is presented. In MCLS, manifold learning is used to transform logical label space to Euclidean label space, and the similarity between samples is constrained by the corresponding numerical labels. The final selection criterion integrates the influence of both the supervision information and local properties of the data. Experimental results demonstrate the effectiveness of the proposed method.
For the past 25 years NIH Image and ImageJ software have been pioneers as open tools for the analysis of scientific images. We discuss the origins, challenges and solutions of these two programs, and ...how their history can serve to advise and inform other software projects.
•Discussed challenges for detection and classification of citrus plant diseases.•Briefly explains recent studies including segmentation and classification.•Compare this review with existing state of ...the arts.•Discussed the advantages and drawbacks of each step with detail.
The citrus plants such as lemons, mandarins, oranges, tangerines, grapefruits, and limes are commonly grown fruits all over the world. The citrus producing companies create a large amount of waste every year whereby 50% of citrus peel is destroyed every year due to different plant diseases. This paper presents a survey on the different methods relevant to citrus plants leaves diseases detection and the classification. The article presents a detailed taxonomy of citrus leaf diseases. Initially, the challenges of each step are discussed in detail, which affects the detection and classification accuracy. In addition, a thorough literature review of automated disease detection and classification methods is presented. To this end, we study different image preprocessing, segmentation, feature extraction, features selection, and classification methods. In addition, also discuss the importance of features extraction and deep learning methods. The survey presents the detailed discussion on studies, outlines their strengths and limitations, and uncovers further research issues. The survey results reveal that the adoption of automated detection and classification methods for citrus plants diseases is still in its infancy. Hence new tools are needed to fully automate the detection and classification processes.
Deep convolutional neural networks have been successfully applied to many image-processing problems in recent works. Popular network architectures often add additional operations and connections to ...the standard architecture to enable training deeper networks. To achieve accurate results in practice, a large number of trainable parameters are often required. Here, we introduce a network architecture based on using dilated convolutions to capture features at different image scales and densely connecting all feature maps with each other. The resulting architecture is able to achieve accurate results with relatively few parameters and consists of a single set of operations, making it easier to implement, train, and apply in practice, and automatically adapts to different problems. We compare results of the proposed network architecture with popular existing architectures for several segmentation problems, showing that the proposed architecture is able to achieve accurate results with fewer parameters, with a reduced risk of overfitting the training data.
•We propose a trainable CNN for weakly illuminated image enhancement.•We propose a Retinex model-based weakly illuminated image synthesis approach.•The proposed method generalizes well to diverse ...weakly illuminated images.
Weak illumination or low light image enhancement as pre-processing is needed in many computer vision tasks. Existing methods show limitations when they are used to enhance weakly illuminated images, especially for the images captured under diverse illumination circumstances. In this letter, we propose a trainable Convolutional Neural Network (CNN) for weakly illuminated image enhancement, namely LightenNet, which takes a weakly illuminated image as input and outputs its illumination map that is subsequently used to obtain the enhanced image based on Retinex model. The proposed method produces visually pleasing results without over or under-enhanced regions. Qualitative and quantitative comparisons are conducted to evaluate the performance of the proposed method. The experimental results demonstrate that the proposed method achieves superior performance than existing methods. Additionally, we propose a new weakly illuminated image synthesis approach, which can be use as a guide for weakly illuminated image enhancement networks training and full-reference image quality assessment.
•A CNN was trained for identification of plant diseases.•Tests considered 12 plant species and 56 diseases.•The effects of using limited datasets were investigated.•The effects of background removal ...were investigated.•Accuracies varied from 60% to 100% depending on the characteristics of each crop.
The problem of automatic recognition of plant diseases has been historically based on conventional machine learning techniques such as Support Vector Machines, Multilayer Perceptron Neural Networks and Decision Trees. However, the prevailing approach has shifted to the application of deep learning concepts, with focus on Convolutional Neural Networks (CNNs). In general, this kind of technique requires large datasets containing a wide variety of conditions to work properly. This is an important limitation, given the many challenges involved in the construction of a suitable image database. In this context, this study investigates how the size and variety of the datasets impact the effectiveness of deep learning techniques applied to plant pathology. This investigation was based on an image database containing 12 plant species, each presenting very different characteristics in terms of number of samples, number of diseases and variety of conditions. Experimental results indicate that while the technical constraints linked to automatic plant disease classification have been largely overcome, the use of limited image datasets for training brings many undesirable consequences that still prevent the effective dissemination of this type of technology.
The Flickr30k dataset has become a standard benchmark for sentence-based image description. This paper presents Flickr30k Entities, which augments the 158k captions from Flickr30k with 244k ...coreference chains, linking mentions of the same entities across different captions for the same image, and associating them with 276k manually annotated bounding boxes. Such annotations are essential for continued progress in automatic image description and grounded language understanding. They enable us to define a new benchmark for localization of textual entity mentions in an image. We present a strong baseline for this task that combines an image-text embedding, detectors for common objects, a color classifier, and a bias towards selecting larger objects. While our baseline rivals in accuracy more complex state-of-the-art models, we show that its gains cannot be easily parlayed into improvements on such tasks as image-sentence retrieval, thus underlining the limitations of current methods and the need for further research.
Existing super-resolution fluorescence microscopes compromise acquisition speed to provide subdiffractive sample information. We report an analog implementation of structured illumination microscopy ...that enables three-dimensional (3D) super-resolution imaging with a lateral resolution of 145 nm and an axial resolution of 350 nm at acquisition speeds up to 100 Hz. By using optical instead of digital image-processing operations, we removed the need to capture, store and combine multiple camera exposures, increasing data acquisition rates 10- to 100-fold over other super-resolution microscopes and acquiring and displaying super-resolution images in real time. Low excitation intensities allow imaging over hundreds of 2D sections, and combined physical and computational sectioning allow similar depth penetration to spinning-disk confocal microscopy. We demonstrate the capability of our system by imaging fine, rapidly moving structures including motor-driven organelles in human lung fibroblasts and the cytoskeleton of flowing blood cells within developing zebrafish embryos.