Shivam K. Thakkar
I am a proactive Data Science Enthusiast working around Machine Learning, Deep Learning, Computer Vision and Natural Language Processing willing to learn every day and enhancing my skill set.
I have completed many projects in the above-mentioned domains, recently I have made a Face Recognition System by replicating Google's FaceNet paper where I worked on Deep Learning models CNN and Siamese Network specifically.
Also, another project I would like to mention is Correlated Words where I used different NLP algorithms to create word embeddings so that correlation between words can be measured simply using cosine similarity. You can consider this a naive approach, Google's Word2Vec and Stanford's GloVe does the same to attain greater accuracy using advanced methods.
Another interesting project where I worked was CardNet where the goal was to build a system that can improve the OCR results of different cards. The output of CardNet can be input to OCR which in turn will extract text. It was able to extract identity, debit, credit, business, etc. cards from an image irrespective of orientation and translation. CardNet is a complex Deep Neural Net architecture which was designed and intensively trained by identity card images that were manually annotated. Along with Deep Learning also Computer Vision algorithms were used for post-processing to improve results.