During my undergrad, I worked as an independent researcher at IUBAT CSE Robot-Vision Lab and also as an undergraduate Research Assistant (RA) for multiple projects at the Miyan Research Institute under the supervision of Prof. Rashedul Islam, Prof. Md. Alomgir Hossain and Prof. Aminun Nahar (from 2017 to 2020).

After my graduation, I was fortunate to get an opportunity to work as a Research Assistant under the guidance of Prof. Dr. Md. Ezharul Islam at Jahangirnagar University for a novel research project of Deep Learning based Computer Vision and Robotics in 2021 which led us winning the Best Paper Award in the International Conference of MIDAS 2021, Springer.

Later on, between 2021 to 2024, I worked as a Researcher and Deep Learning Algorithm Developer in Cisscom (United States), KaleidoSoft (Croatia) and Vinacts (South Korea). Currently, I am working as a Research Assistant (RA) with Prof. Dr. Humayun Kabir from the Department of Integrated System Engineering at Inha University, South Korea. My research field relates with developing state of the art efficient algorithms for Robot Vision capabilities and medical image processing; occlusion aware object tracking and frame interpolation mechanism.

In this research, we initiated a unique and cutting-edge backbone neural network for the conventional YOLO algorithm which we named as SBHK-Net. The network boosted up the performance of the existing YOLO algorithm drastically which manifests a strong potential of improving tracking and recognition accuracies of other conventional algorithms in the robot vision industry as well. It has the greatest accuracy 59.2% AP among all known real-time object detectors with 30 FPS or above on GPU RTX3060, and it outperforms all other known object detectors in the range of 5 FPS to 160 FPS. We used YOLOv7 as our reference point for the core research. The transformer-based detector SWINL Cascade-Mask R-CNN (9.2 FPS A100, 53.9% AP) and the convolutional detector ConvNeXt-XL Cascade-Mask R-CNN (8.6 FPS A100, 55.2% AP) are both outperformed by the SBHK-Net core object detector (56 FPS RTX3060, 56.4% AP) in terms of speed and accuracy, respectively.

We focused upon the deep learning approach to classify the normal and abnormal breast according to the medical imaging from the MIAS dataset of Mammograms and Pixel Intensity. The Convolution Neural Network (CNN) alongside ResNet, AmoebaNet and EfficientNet have been used for the detection with 330 mammograms in which 194 images are normal and 136 are having the identification of abnormal breasts. The accuracy of the entire experimental results was carrying the torch of potential legacy of deep learning in the medical imaging arena. The research is ongoing for the further development and optimization of CNN, AmoebaNet-C and EfficientNet architecture for the Pixel Intensity with higher accuracy, proper segmentation and masking.

In this research, we focused mostly on our rigorous novel implementations and iterative result analysis of different cutting-edge modified versions of EfficientNet architectures namely EfficientNet-V1 (b0-b7) and EfficientNet-V2 (b0-b3) with ultrasound image, named as CEIMVEN. We utilized transfer learning approach here for using the pre-trained models of EfficientNet versions. The approximate testing accuracies we got from the modified versions of EfficientNet-V1 (b0- 99.15%, b1- 98.58%, b2- 98.43%, b3- 98.01%, b4- 98.86%, b5- 97.72%, b6- 97.72%, b7- 98.72%) and EfficientNet-V2 (b0- 99.29%, b1- 99.01%, b2- 98.72%, b3- 99.43%) are showing very bright future and strong potentials of deep learning approach for the successful detection and classification of breast cancers from the ultrasound images at a very early stage.

Nano Rover is a significant approach of cost-efficient surveillance and reconnaissance robot which is fully functional and cost-efficient at the same time. It features the service of active reconnaissance mode with LIDAR sensor, location tracking with GPS Neo 6M module, visual information collection, person detection, weapons detection and identification, gender and age prediction of the hostile and other artificial threat detection, etc. Remote navigation plays as the core controlling system of the robot which is also modifiable through replacement with Internet and satellite navigation system. We modified the conventional Inception-Net architecture with a better hyper-parameter tuning for the successful execution of image processing tasks with a better level of accuracy so far.

This robotics research project proposes a solution which appears to be a full-fledged Bluetooth controlled Submarine prototype with a sensory chipboard attached inside its endo-skeleton which contains multiple sensors like DHT11 temperature-humidity, dust, CO2 and YL69 pH sensors. The sensory data provides the information of underwater whether the naval environment is habitable for the marine biological species or not, under the terrible effect of global climate change. The submarine prototype is fully functional in the surface and underwater scenario which contains a very unique mechanical design and circuitry with an exceptional sensor data streaming capability which can be used by marine biological researchers and oceanographers professionally as a full-fledged marine ecosystem monitoring device.

In this research, we tried to represent an optimized implementation of different chatbot algorithmic approaches within a single University Automation query platform where the previous chatbots were developed concerning the utility areas of agriculture, economics, medical science with disease prediction, admission system and tourism. They were manifested with Deep Neural Networks like CNN, LSTM and NLP with different environments and approaches. Our primary concern was to combine and optimize all of the approaches in a single platform of a University Automation System as much as possible having Keras as backend. The accuracy rate seems to be very promising through the amalgamation of CNN layers and LSTM with NLP features.

In the rural world of medical services, we generally notice a lot of havoc which generally happens in the hospitals, clinics and related other medical centers. The conditions of Intensive Care Units (ICU) of the rural areas are quite intolerable because of the lack of qualified nurses. Doctors generally prescribe multiple injections for a single patient for each day. Nurses are responsible for the injection process but unfortunately they fail to perform the injection process very often in proper prescribed time in proper amount. This malpractice of treatment quite often results in the terrible sufferings of the ICU patients and sometimes a few patients even die. This study aims to minimize the hazard at the highest accuracy level possible. The research relates to the functionality of Real Time Clock (RTC) which provides the activation of automated time system and triggers the microcontroller's machinery to act according to the time. According to the doctor's prescribed time, injection's medicine will be flowing inside the pipelines and will be injected to the body of the patient through a cannula. In this study, Arduino microcontroller, RTC DS3231 time module, HX 711 weight sensors, relay modules, hydraulic pump motors, wifi shield, resistors, MOSFET and breadboard have been used. Following the prescribed time of doctor, the RTC module programs the time for the activation of the microcontroller. The microcontroller activates the hydraulic pump motors following the programmed times. The pump motors then create a vacuum environment inside the pipeline and pass the medicine fluid for injection inside the patient's body through the multiple channel-single cannula. This is consisted of three units: electrical circuitry unit, mechanical unit and timer program with reprogram process unit. The research were carried out through the tests of each and every units to verify that they were working precisely and were manifesting the expected outputs.