Innovated a novel grid-based clustering algorithm, which reduces the number of comparisons for forming clusters exponentially, resulting in reduction of the overall time complexity to 𝑛^(3/2) better than 𝑛^2 complexity of the traditional DBSCAN algorithm. Also, we acquired a significant computational speed up-to 58% over other improvements of the DBSCAN algorithm while maintaining 100% accuracy.
Engineered the concept of Multi-view Face Detection and Tagging using Convolutional Neural Networks (CNN) - identifying faces from an image and provide labels to the detected faces using the Tensor-flow framework and Caffe library.Also, using this approach, we acquired an overall accuracy of 85% for facial recognition.
Constructed a low-cost wireless sensor-based end-to-end system using Ultrasonic sensors, Arduino Uno R3, GPS module, Gyroscope and Accelerometer. Further, the location of detected potholes are notified to the appropriate government bodies using this system.