Pervasive computing has made almost every device we see today to be communicated and function in collaboration with one another. Since the portable devices have become a part of our everyday life, people are more involved in a pervasive computing environment. They engage with many computational devices simultaneously without knowing the availability of their existence. The current world is being filled with more and more smart environments. These smart environments make them to be attracted towards the new technological emergence in the field of pervasive computing. Various researches are being carried out to improve the smart environment and their applications. Middleware plays a vital role in building the pervasive applications. The pervasive devices act based on the context of the situation, that is, they do their actions according to the environment of the application. They react to the situations smartly as they can take their own decisions based on the context developed for that specific application. Most of the pervasive applications were using its own middleware that is specific towards their need. As today, most of the applications are using their own middleware with their specific requirement, which leads us to unearth out their common features and their scope of using it. In this paper, a survey on the various hybrid aspects of the different context-aware middleware has been done. This middleware is classified based on service, context, and device aspects. Merits and demerits are identified from the existing smart environments, and future perspective of their development such as generic context-aware middleware need has been discussed.

India is one of the leading productions of Paddy. Compared to previous year Gross Domestic Product (GDP) Export rate of Paddy in the year 2021 has increased to around 33%. Paddy is the Major food production crop in India. Every crop is prone to many diseases throughout their lifespan. The disease can affect the crop at any stage of their growing phase. Early detection of disease is the only solution to reduce the damage. Early detection may reduce the damage caused and increase the quality as well as quantity of Production. Major disease which causes more damage in paddy production is Rice Blast, Brown Spot, Sheath Blight, Sheath Rot and False Smut. Early detection of these diseases can reduce the damage and increase the production value. Recent technology of computer vision and by using Deep learning model can accurately predict and diagnose the early symptom of diseases. We used Convolutional Neural Network classifier of deep learning model to predict the early symptom of disease in paddy. We compared four main classifier VGG16, VGG19, Inception-V3 and ResNet50, among these four Inception-V3 achieved a highest accuracy of 95.3%.

Weed is an unwanted plant which is grown in agriculture land. The land which is not cultivated will be fully covered by Weeds. Management of weed is the major concern for farmer because the weed will reduce the crop production quantity. There are many methods to control the weeds, one of those methods is manual plucking which is expensive because it takes more time, consumes human work. Second is by applying any chemicals suggested by external experts. This may cause damage to the crop which is cultivated. Identifying weeds in early stage of crop growth and destroying them through proper method is most important for increasing the crop production. We proposed an efficient method for identifying and classifying weed in paddy field by using Deep learning-based computer vision techniques. We applied Semantic Segmentation model for classifying weeds in agriculture land. We trained our model with SegNet with different batch size of 16,32,64 and obtained a highest accuracy of 94.223 for dropout value 0.1 and batch size set to 32.

Deep learning is reaching new heights as a result of its cutting-edge performance in a variety of fields, including computer vision, natural language processing, time series analysis, and healthcare. Deep learning is implemented using batch and stochastic gradient descent methods, as well as a few optimizers; however, this led to subpar model performance. However, there is now a lot of effort being done to improve deep learning’s performance using gradient optimization methods. The suggested work analyses convolutional neural networks (CNN) and deep neural networks (DNN) using several cutting-edge optimizers to enhance the performance of architectures. This work uses specific optimizers (SGD, RMSprop, Adam, Adadelta, etc.) to enhance the performance of designs using different types of datasets for result matching. A thorough report on the optimizers’ performance across a variety of architectures and datasets finishes the study effort. This research will be helpful to researchers in developing their framework and appropriate architecture optimizers. The proposed work involves eight new optimizers using four CNN and DNN architectures. The experimental results exploit breakthrough results for improving the efficiency of CNN and DNN architectures using various datasets.