This research presents a completely new paradigm for seizure onset detection automation that is based on electroencephalography (EEG). In order to process the data, we used the CHB-MIT scalp EEG dataset. Data augmentation is utilised so that the dataset can be balanced out. We had to increase the size of the ictal data since, in compared to the interictal records, they were far too little. In order to deconstruct the EEG signal, one of the following techniques may be utilised: a discrete wavelet transform (DWT)-Daubechies Wavelet4(db4), a continuous wavelet transform (CWT), a wavelet packet decomposition (WPD), or a continuous wavelet transform (DWT)-Daubechies Wavelet4(dbS).It was determined that the VGG16 pre-trained model was the best option for automatically classifying 2D scalogram images into seizure and non-epileptic seizure states. The performance of four wavelet-based features was tested, and CWT-based features outperformed all others with an average accuracy of 97%.
Medico-visual processing is a rapidly expanding and significant field right now. A number of modules, including X-ray, ultrasound, MRI, CAT, PET, SPECT, and BIOPSY, are used to get intrinsic uniformity and motifs utilising magnetic and radio waves. The excellent intention and coherence of MRI, a3D clinical imaging technique, are thought to make it the most suitable for image preprocessing. We can now identify even the tiniest physical flaws in people because of these technologies. To reliably and usefully retrieve information from images with the fewest possible mistakes is the primary objective of medico-visual processing. Because it produces images with higher resolution and better quality than those produced by other imaging technologies. The technique of detecting brain tumours using MRI pictures is divided into four stages: preprocessing, image segmentation, feature extraction, and image classification. Because of the brain’s intricacy, identifying brain tumours using MRI scans is a difficult process. Image segmentation is possible after preprocessing MRI images. These tumours can be segmented using several image segmentation algorithms such as SVM, CNN, FCM, fuzzy K-mean clustering, DL, U-Net, PDCR, UAFR, Swarm Learning, GFR, LFR, KL, HMRF, FCEM, IFCM, MISSFormer, and image classifiers such as lazy IBK, SVM and GZSL.
A field that is constantly gaining popularity and has an increasing number of users is online shopping. The ease of shopping from the comfort of our homes along with the wide variety of choices at our disposal makes it the preferred choice of most shoppers. However, online shopping has a big problem that has not been addressed completely. A major problem in online shopping is the fact that picking the correct size or deciding if a particular item would suit a person. This makes online shopping a big hassle especially when we have to return certain items. The Computer Vision based shopping assistant helps provide a virtual tool to view the demos of the clothes that we wish to buy. It has an extensive catalogue of clothes to shop from along with virtual demo options so that when we buy a piece of clothing, we know exactly what we are getting and its size and fitting.
The study of changes in Land-Use and Land-Cover (LULC) helps in planning and management of urban expansion for an area. The attempt made to study the changes of LULC for Pune District of Maharashtra using LANDSAT images of 2001, 2009, 2014, and 2019 is presented in this article. Maximum Likelihood Algorithm is used for classifying the image into four classes; Built-Up, Vegetation, Water, and Barren. Performance of classification is verified manually taking reference points. The result shows, built-up land was increased from 2.24 % in the year 2001 to 11.19 % in 2019, vegetation land initially increased from 30.34 % in 2001 to 35.51 % in 2009 later decreased to 27.51 % in 2019. Similarly, water area increased slightly from 2.17 % in 2001 to 2.96 % in 2009 later decreased to 2.41 % in 2019. Barren land decreased from 65.25 % in 2001 to 57.27 % in 2009 then increased to 61.82 % in 2014 and finally decreased to 58.89 % in 2019.
CNNs and Edge computing have proliferated in the last decade. While these technologies are stupendous in their own right, new horizons can be opened up if they were to be combined. Deep and complex CNNs, which are ones that produce groundbreaking accuracies, are not feasible for deployment in a resource constraint environment. The vanguard in model compression is an innovative technique known as Knowledge Distillation. This paper proposes a way to perform knowledge distillation by means of Attention Maps. The generation of attention maps from features is achieved through the use of ATCam. Not only does this method effectively distills knowledge it also nullifies another flaw of CNNs-black-boxiness. This may lead to leapfrogs in Mobile Machine Aided Medical Diagnosis by deploying interpretable complex CNN in small devices. Three ways to combine the feature maps were experimented with in this research work, which are Feature Map Accumulation(FMA), Maximum Intensity Projection(MIPs) and Saliency Maps. The performance of the proposed approaches are tested on PADUFES-20 dataset and all the three approaches provided significant boost in the performance over the traditional transfer learning methods. The best of these methods - saliency maps - gains a accuracy boost of 9.56% over the generic methods of knowledge transfer. Moreover, to one’s surprise, the resultant model not only matches the teacher model in accuracy, it exceeds it by 1.52% while being only 43.2% of original teacher model’s size.
