Research interests
The main theme of my research is to use machine learning and deep learning model which can have impact on real-life situations. I love to work on both feature extraction level and model building level. As the size of data is increasing day-by-day, extracting important features is very crucial for better optimization of models. I am interested in working with medical image data, healthcare data, object detection from traffic images and thermal image data analysis.
Research Projects
- Lung Cancer detection: This research project is based on a kaggle competition for improving the accuracy of lung cancer from chest X-ray reports. The goal of this project was to identify cancer images with high accuracy from X-rays. The first challenge in this project was to extract features without any loss. The dataset consists of nearly 50k cancer and non-cancer augmented images. I have used SVM, Decision Tree and Deep Learning models to get the optimal result.
- Malaria Infected Cell Classification: Malaria is one of the most devastating endemic diseases in Africa and Asia. Thousands of lives are lost each year because of the spread out of malaria parasite. The high infection rate is aided by a lack of experience in utilizing lab equipment accurately and prompt diagnoses. I have used customized CNN with padding trick in feature mapping level for capturing as much information as we can. To measure the efficiency of the proposed model, I have compared our model with other state-of-art models which performed on same NIH malaria dataset. The proposed model is evaluated using standard performance matrices such as Accuracy, Precision, Recall, F1 Score.
- Face Mask Detection: In this Covid-19 pandemic period, face mask announced mandatory by different countries of the world. But very few people from third world countries obey this rule properly. It is also not possible to monitor every person in the street for face mask checking. After the start of pandemic there have been lost of work done on face mask detection from daylight images. But there have been no research on the night vision images. I have worked on infrared and thermal image data for face mask detection after daylight. I have used hybrid deep learning model for this research.
Academic Projects
- Weather Forecast Application: This is a weather forecast application that can show the weather forecast of the current and for the next five days. The weather forecast is found based on the location of the cell phone.
Tools: Java, Android, REST API, JSON.
- Smart Attendance System: This is an android based attendance system application. All the data stored in a central server, so any change made will be present in any device without any further maintenance.
Tools: HTML, CSS, PHP, JavaScript, MySQL, Bootstrap, XML, Java, JSON, REST API, Glide.