Rahul Mangalampalli

- Used multinomial regression method to predict house prices. - Used mix of methods like Naive Bayes, KNN, Random forest, XGBoost , AdaBoost Ensembling to predict credit card fraud and potential customers for a cellular services company.

CNN based projects: - Trained multiple state-of-the-art models in Pytorch, Caffe, Tensorflow with peculiar datasets. Achieved a uptill 80%-90% accuracy on every model. - Designed a pipeline to train StyleGAN that uses neural style transfer to generate images in the same probability distribution of a certain dataset. - Trained and deployed MobileNetSSD for detecting professional ID cards which not easily visible. - Added improvised average pooling layer to Caffe framework. - Converted teachable machine's MobileNetV2 model into Caffe. Designed a C++ API with OpenCV for using the teachable machine on Raspberry Pi with armv8 architecture. OpenCV based projects: - Designed End-to-End application to detect cellular organsims in an image using improvised hybrids methods based on Watershed Algorithm - Designed a method to detect number plates and the characters present in it using a k-mean color clustering method. - Designed a method to detect aadhar cards and pan cards using template matching and contrastive background manipulation techniques.