Khalequzzaman Likhon | Computer Vision Enthusiast

Hi, I'm Khalequzzaman Likhon.

A

I am fascinated with the AI driven applications which are meant to reduce human labor. I believe that this field has immense challenges and opportunities to keep me excited for a lifetime. I love to play football, play guitar, and travel.

About

I hold a Bachelor of Science in Computer Science and Engineering from Ahsanullah University of Science and Technology, where I gained experience with technologies such as Java, Python, MySQL, and MVC architecture. My academic journey, particularly through courses in Artificial Intelligence, Machine Learning, and Digital Image Processing, sparked my interest in deep learning and computer vision. I have worked on projects involving object detection, image classification, segmentation, and sentiment analysis, with a focus on developing innovative techniques for image analysis and visual understanding. My primary research interest lies in computer vision, exploring the intersection of attention mechanisms, multi-modal learning, and self-supervised learning to enhance model performance. I am also intrigued by the convergence of natural language processing (NLP) and large language models (LLMs) with computer vision, particularly for applications like vision-language models and cross-modal learning. Feel free to reach out if you're interested in collaborating: khalequzzamanlikhon@gmail.com

Research

Design of an Arrhythmia Classification Algorithm Using 2-D Convolutional Neural Networks.

Experimented on the MIT-BIH Arrhythmia database which is the first available data on the basis of ECG data from the Massachusetts Institute of Technology–Beth Israel Hospital (MIT–BIH). The automated systems can be adapted as a tool for screening the ECG images to classify arrhythmia classification. This could play a vital role not only for the patients but can also assist the doctors.

Research Interest

Deep Lerarning, Computer Vision

Image Processing, Object Detection, Activity Recognition, Object Recognition, 3D pose estimation, Motion Estimation, video surveillance and security (real-time video analysis, object tracking, and anomaly detection in surveillance systems), and video analytics.

Projects

                
Pneumonia Detection
                  Extracting features using bothe densenet121 and efficientnet and applying attention ensemble
              
AccomplishmentsTools: Keras, TensorFlow, openCV, Transfer Learning, numpy, pandas, matplotlib 
The trained model can detect Pneumonia in CXR with 95% accuracy

Object Detection
                  Pretrained model trained on COCO dataset is used here to train the object detection model on PASCAL VOC-2012
              
AccomplishmentsTools: pyTorch, openCV, Transfer Lerarning, numpy, pandas, matplotlib 
The trained model can detect objects within particular 20 calsses

Image Classification on CIFAR-10 Dataset
                  Custom Convolutional Neural Networks model is trained on CIFAR-10 using different deep learning techniques
              
AccomplishmentsTools: Keras, TensorFlow, Python, numpy, pandas, seaborn, matplotlib
The trained model can classify new images.

Survival Prediction in Titanic
                  Logistic Regression algorithm on Titanic Dataset
                
AccomplishmentsTools:Python, Numpy, Pandas, matplotlib, Streamlit
The model can predict percentage of Survival of a person in titanic based on the 
                    features in the dataset.
                  
Diabetes Detection
                  Naive Bayes' theorem in predicting probability for unknown class
                
AccomplishmentsTools: Python, Pandas, Numpy, matplotlib, Streamlit
The model can calssify whether a person do have diabetes or not based on different features.

Char Level Language Model
                  char level language model using Recurrent Neural Networks(RNN)
                
AccomplishmentsTools: Python, RNN, Numpy, Streamlit
The model can generate text only similar type of the text it was trained with