Projects
Image Classification of Stroke Blood Clot Origin
Objective: To analyze pathology images of blood clots and classify the etiology to either CE (i.e., originating from the heart) or LAA (i.e., originating from the plaque in the inner lining of an artery).
Used DenseNet and ResNet network with transfer learning to classify blood clot images to detect their origin. Achieved an accuracy of 72.4% and 73.75% with an F1 score of 0.701 and 0.765 respectively Project link
Object Detection using YOLOv2
Incorporated the YOLOv2 object detection algorithm with pre-trained weights to accurately and efficiently detect various objects in real-world images, demonstrating expertise in key deep learning concepts such as CNN, anchor boxes, and non-max suppression.
SPEAR: Soft Robotic EMG Assisted Rehabilitation
A bio-inspired solution for foot rehabilitation of stroke patients suffering from temporary paralysis.
Constructed a pneumatic circuit with solenoid valves for movement and pressure control of the actuators
Obtained Electromyographic signals from muscles for dorsiflexion and plantarflexion movement of the foot from calf muscles.