Imitation Learning, RESNet, CARLA simulator
In autonomous driving, exact precision is indispensable for an agent to navigate safely. So we created an imitation learning based model which uses an expert’s past driving data, collected by Carla simulator and imitates those commands for the exact or similar situation.
Vehicle Detection with Yolov5 and Efficientdet
For the DhakaAI Vehicle Detection Challenge, our project focused on leveraging benchmark object detection models, specifically YOLOv5 and EfficientDet, to accurately identify and classify vehicles in complex urban environments. We enhanced model performance through comprehensive data augmentation techniques, which improved the model's generalization and robustness. Additionally, we implemented an ensembling strategy, combining the strengths of multiple models to achieve superior accuracy. These efforts resulted in our team consistently ranking in the top 10 for both the first and second leaderboards, demonstrating the effectiveness of our approach.
DjangoREST, ReactJS, MySQL
Built an apartment management system for automate the functionalities of an apartment e.g committee election, polling for decision, payment and tracking all the finances of the apartment.
Microcontroller: Atmega32
We have made a mini version of piano using the Atmega32 Microcontroller which is able to play different musical notes. We've been a bit ambitious here by making this piano play according to the finger gestures made by the users. Also the users can record and save the tune they play and playback it anytime they want.