IEEE International Conference Image Processing, Applications and Systems 

Abstract: Recognition of grocery products in store shelves poses peculiar challenges. Firstly, the task mandates the recognition of an extremely high number of different items, in the order of several thousands for medium-small shops, with many of them featuring small inter and intra class variability. Then, available product databases usually include just one or a few studio-quality images per product (referred to herein as reference images), whilst at test time recognition is performed on pictures displaying a portion of a shelf containing several products and taken in the store by cheap cameras (referred to herein as query images). Moreover, as the items on sale in a store as well as their appearance change frequently overtime, a practical recognition system should handle seamlessly new products/packages. We developed a deep learning based pipeline to solve this task. First we deploy state of the art object detectors to obtain an initial product-agnostic item detection, then, we pursue product recognition through a similarity search between global descriptors computed on reference and cropped query images. To maximize performance, we learn an ad-hoc global descriptor by a CNN trained on reference images based on an image embedding loss. We have tested our pipeline on the standard grocery product [1] dataset and improved the currents state of the art. While computationally expensive at training time our system turn out not only accurate but also quite fast at test time.

Abstract: Convolutional Neural Networks (CNNs) have the potential to assist medical doctors in diagnosis and treatment stage. This paper has been prepared to help dermatologists by presenting (i) fundamental information on deep learning and CNNs, and (ii) applications of CNN s for skin diseases classification. Also, this work shows that although CNN based methods have a strong potential for automated diagnosis, further researches and new techniques are still required in image processing and pattern recognition area to provide diagnosis of dermatological diseases with higher performances. In this work, these two groups of applications of CNN s in dermatology have been handled: (i) disease classification from medical images (e.g., dermoscopy and pathological images); (ii) disease classification from digital photographs. Therefore, important contributions of this work are two-fold: First, main concepts of deep learning and CNNs are presented, which will be helpful for dermatologists to understand and follow up CNN based computerized methods. Second, the state-of-the-art applications developed for lesion classifications from medical images and colored photographs are presented. Also, disadvantages or limitations of these applications are explained. In addition, this paper indicates shortage of desktop applications developed for other dermatological diseases except skin cancer.

Abstract: Coconut production is the most important and one of the main sources of income in the Sri Lankan economy. The recent time it has been observed that most of the coconut trees are affected by the diseases which gradually reduces the strength and production of coconut. Most of the tree leaves are affected by pest diseases and nutrient deficiency. Our main intensive is to enhance the livelihood of coconut leaves and identify the diseases at the early stage so that farmers get more benefits from coconut production. This paper proposes the detection of pest attack and nutrient deficiency in the coconut leaves and analysis of the diseases. Coconut leaves monitorization has been taken place after the use of pesticides and fertilizer with the help of the finest machine learning and image processing techniques. Rather than human experts, automatic recognition will be beneficial and the fastest approach to identify the diseases in the coconut leaves very efficiently. Thus, in this project, we developed an android mobile application to identify the pests by their food behaviors, pest diseases and the nutrition deficiencies in the coconut trees. As an initial step, all datasets for image processing technology met pre-processing steps such as converting RGB to greyscale, filtering, resizing, horizontal flip and vertical flip. After completing the above steps, the classification was performed by analyzing several algorithms in the literature review. SVM and CNN were chosen as the best and appropriate classifier with 93.54% and 93.72% of accuracy respectively. The outcome of this project will help the farmers to increase the coconut production and undoubtedly will make a revolution in the agriculture sector.

Abstract: Performance achievable by modern deep learning approaches are directly related to the amount of data used at training time. Unfortunately, the annotation process is notoriously tedious and expensive, especially for pixel-wise tasks like semantic segmentation. Recent works have proposed to rely on synthetically generated imagery to ease the training set creation. However, models trained on these kind of data usually under-perform on real images due to the well known issue of domain shift. We address this problem by learning a domain-to-domain image translation GAN to shrink the gap between real and synthetic images. Peculiarly to our method, we introduce semantic constraints into the generation process to both avoid artifacts and guide the synthesis. To prove the effectiveness of our proposal, we show how a semantic segmentation CNN trained on images from the synthetic GTA dataset adapted by our method can improve performance by more than 16% mIoU with respect to the same model trained on synthetic images.