1. Anshuman Singh
Personal Website-rimijoker.github.io
as4917@srmist.edu.in | 6387520131
github.com/rimijoker | linkedin.com/in/iamanshumansingh
EDUCATION
SRM INSTITUTE
OF SCIENCE AND TECHNOLOGY
(KTR)
B.TECH (COMPUTER SCIENCE
ENGINEERING)
2022 (expected) | Chennai, IND
CGPA-8.6/10
ST. ANJANI'S PUBLIC
SCHOOL
CBSE CLASS XII
2017 | Lucknow, IND
COURSEWORK
UNDERGRADUATE
Calculus and Linear Algebra
Programming for Problem Solving
Data Structures and Algorithms
Object Oriented Design and Programming
Management Principles for Engineers
Entrepreneurial Skill Development
MOOCS
Deep Learning Specialization
(5 Courses)(deeplearning.ai)
Issuing Organization: Coursera
TensorFlow in Practice Specialization
(4 Courses) (deeplearning.ai)
Issuing Organization: Coursera
Machine Learning
(Stanford University)
Issuing Organization: Coursera
SKILLS
NATURAL LANGUAGE
PROCESSING:
• RNNs• LSTMs •GRUs•Bag of words
•GLoVe Embeddings •Word2Vec
• N-grams
COMPUTER VISION:
• GANs • RCNNs • YOLO • CNNs
MACHINE LEARNING:
• RandomForest • GradientBoosting
• K-Means • kNN • SVM
LIBRARIES:
• Tensorflow • Keras • PyTorch • fastai
• Sci-kit Learn • nltk• XGBoost • Numpy
• Pandas • Matplotlib • Plotly • Selenium
LANGUAGES:
• Python • C++
PROJECTS
COLORIZATION USING GANS AND U-NET(GITHUB) | GANS,
U-NET, CNNS, FASTAI, PYTORCH
•Implemented a U-Net to colorize black and white images into coloured
images.
•The U-Net was built with the fastai framework which runs on top of PyTroch.
•The encoder part of the U-Net was ResNet-18 and for the decoder part the
convolution layers and pooling layers were replace with
transpose-convolution layers and un-pooling layers.
•The loss function was changed to perceptual loss/feature loss.
•The U-Net was per-trained as a regression task with input as grayscale
images and target as coloured images.
•Then the Discriminator was per-trained using the generated images from
the U-Net and the original coloured images.
•Then the U-Net was trained as a GAN.
SHAKESPEARE-GENERATOR(GITHUB) | LSTM, TENSORFLOW,
KERAS
•Implemented a LSTM with with Keras using Tensorflow as backend to
generate text like William Shakespeare.
GENERATING HUMAN FACES WITH GANS(GITHUB) | GANS,
CNNS, TENSORFLOW, KERAS
•Implemented a GAN to generate new human faces using the LWF faces
dataset in Keras using Tensorflow as backend.
•Random Noise was fed into a CNN(Generator) which was trained against
13,000 images of human faces with the help of a Discriminator.
MALARIA PREDICTION USING CELL IMAGES | TRANSFER
LEARNING, CNNS, TENSORFLOW, KERAS
•Implemented a CNN on NIH malaria dataset to classify cell images into
Infected and Uninfected.
•Dense-Net121 architecture with pre-trained image-net weights was used
for the classification and an accuracy of 0.94 was achieved.
CATS VS DOGS | CNNS, TENSORFLOW, KERAS
•Implemented a CNN with with Keras using Tensorflow as backend to classify
images of dogs and cats.
EXPERIENCE
NEXT TECH LAB | MACHINE LEARNING PRACTITIONER
MEMEBER IN MCCARTHY LAB
July 2019 - Present | Chennai, IND
• India’s first multi-disciplinary undergraduate student-led research lab.
• Actively involved in the field of Artificial Intelligence.
• Helping students new to the field by providing a structured learning path.
VOLUNTEER
NSS | NATIONAL SERVICE SCHEME