One of our projects includes extracting stock market data of different tickers and visualizing them to find trends. Visualization parameters were stock price, market cap, gross profit, and revenue, etc. on different time periods of daily, monthly, quarterly, and yearly. Python was used as the language to pre-process and visualize the data
On basis of changes over specific periods of time, companies were shortlisted and a qualitative analysis on those companies was carried out to see why the stock prices drops or rise i.e. in case of McDonald’s, stock price a highly related with opening and closing of new branches.
This project includes converting voice to text using google API in real-time and applying analytics on the text i.e. words per minute, frequency of top 20 words and repetition of words, etc. A simple web page was designed to visualize the output of python based voice analytics.
COVID-19 Classification and Detection
This was a research project in which the task was to classify COVID-19, normal, and pneumonia class. Pre-trained CNN models like ResNet-50, NasNetLarge, and VGG-16 were used for classification after fine-tuning. For localization of the affected area, a technique called grad-cam was used. Python was used as the language used for coding.
Health Monitoring Literature Review
This was a research project to highlight the previous work done in the health monitoring domain. A literature review was done in the following domains:
Disease and pain detection through video
Emotion Recognition using face video
Fatigue detection using face
Health monitoring through sensors
Heart rate estimate from face
Seizure detection using EEG
Seizure detection using face
Moth-Eye SEM Images Analytics
This project investigates one application to the analysis of image data for speciï¬c nonuniform spatial features. Spatial cross-correlation, spectral cross-correlation and convolution are investigated as possible image processing techniques. Further real-world spatial distances are determined from images from depth maps. These techniques are combined with applying statistical. Find statistical information about moth-eye SEM images. The defect quality and other properties are determined using computer vision algorithms and feature extraction. The ï¬nal program gives an output of statistics for the moth-eye SEM image. These three experiments were conducted:
Experiment: Cross-correlation in 1D and 2D
Experiment: Patterns in Stereovision
Application: Moth-eye SEM image analysis
Electrical Grid Stability Analysis Using Machine Learning
Electrical grid stability has become a major issue with the uprise of renewable energy sources as it is difficult to keep the steady flow in the bi-directional systems. To check the stability of the grid some parameters are suggested on which datasets are formed. One of the Dataset is "Electrical Grid Stability Simulated Data Set" by UCI. It contains different parameters on which the stability of the Grid is determined. On basis of these parameters, features have been identified that have the same data across 90% of the rows. Different Machine Learning approaches i.e. Neural Networks, Random Forest, SVM (State vector Machine), KNN ( KNearest Neighbours ), Decision Trees, etc. have been implemented and compared in this paper and the best model has been chosen to check the stability of the grid.
Analysis to Target Existing Customers to Buy Services
This was a market project in which we have the data of customers related to a specific business. This was a uni-class classification problem where we grouped our customers who have already bought the services and applying the classifying model to customers who have not bought the services yet. In the end, we got the list of customers who are more likely to buy a specific service from all the customers who have not bought the service yet. Python was used to implement data pre-processing and applying machine learning algorithms.
Review of Bio-Metric Recognition
This was a research project in which a review of biometric recognition systems was done. Biometric recognition has been a very important way of surveillance for a very long time now. There has been a very vast work present in the field of both multi-model and uni-model recognition systems using deep learning along with preprocessing done through Generative Adversarial Networks (GANs). we looked at the latest trends and work present until now. We discussed all the strategies and datasets that are used mostly along with accuracies achieved by those strategies. The project aims to give an overview of the field of biometric recognition through deep learning.
COVID-19 Normal and 14 Other Chest Pathology X-Ray Classification
COVID-19 has spread at a very fast rate and it is important to build a system that can detect it in order to help overwhelmed health care system. The strength of deep learning techniques is used in many research studies of chest-related diseases. Although some of these researchers used state-of-the-art techniques and was able to provide promising results, these methods do not provide much benefit if it can only identify one type of disease without identifying the rest. In order to deal with this problem, a novel pipeline was introduced for the detection of COVID-19 along with other chest diseases from X-ray images. This pipeline reduces the burden of classifying a large number of classes on a single network. Lungs and heart are segmented from the whole X-ray image and passed onto the first classifier that checks if the X-ray is normal, COVID-19 affected or belongs to other chest X-ray disease. If the case is neither COVID-19 nor normal then the second classifier comes into action and classifies the image as one of the other 14 diseases. With our new pipeline, we show how our model makes use of the CNN state-of-the-art deep neural networks to achieve COVID-19 classification accuracy with 14 other chest diseases and normal X-ray images that is competitive with present state-of-the-art models.
Low Cost Assembly Design of Unmanned Underwater Vehicle (UUV)
Unmanned Underwater Vehicle (UUV) drives underwater without requiring any real-time input from a human operator. These robotic vehicles are mostly used to explore the underwater world. These vehicles can be designed to perform underwater specific tasks like the exploration of the underwater world and spying etc. The readily available UUV’s present in the market is of high cost ranging from 2000$ to 3000$. To reduce the production cost of a UUV, many pieces of research are in process worldwide. A lot of physics concepts have to be used while designing such vehicles. Image Processing is an essential feature of UUV based on which it performs underwater tasks. There are many constraints in performing underwater image processing as the turbidity of the water changes when one goes deep in the water. It mostly suffers scattering, limited range, low contrast, and most importantly blur in the images that create problems for UUV to perform its tasks efficiently. We propose a basic software, electrical, and mechanical hardware design for UUV. PVC pipes were used for designing the structure of UUV, Raspberry pi acts as the processor of the UUV, and bilge pumps act as propellers. Our design is much simpler than the readily available UUV’s in the market. It is a low-cost UUV that performs specific tasks like goal post-detection, color detection, and obstacle detection based on template matching. It is developed for under 300$. One can add more features to a UUV to perform underwater photography and observation of marine life. The efficiency of the UUV system and battery life is enhanced for better functionality. A special waterproof container is being developed and attached to the UUV structure for waterproofing of the whole circuitry.
Real-Time shallow water image retrieval and enhancement for low-cost unmanned underwater vehicle using Raspberry Pi
Unmanned Underwater Vehicles (UUV) operated underwater without any human interference. UUVs can be used for many purposes i.e. exploration or spying. They have several designs and each one can be used for a specific purpose. The main problem that limits the functionality of UUV is poor image quality. The reason behind this poor image quality is scattering and wavelength absorption. The purpose of this research was to propose an electrical assembly design of low-cost UUV. In order to carry out this work use of Raspberry Pi is proposed which is a small computer on its own. It can be used to enhance image quality and detect objects along with color detection. For controlling the propellers in accordance with any hindrance, pulse width modulation (PWM) GPIO pins are used. GPIO pins are also used to attach sensors. For enhancement of images python 3 was used and different image enhancement methods were tested on basis of execution time per frame. In the end best model is selected which is more suitable on basis of execution time and image enhancement. In order to transmit the images from UUV to the base computer wi-fi module of Raspberry pi is used.