Supriya Mandal | Software Engineer

Hi, I'm Supriya Mandal.

A

I'm a Software Engineer in a SaaS company. I did my postgrad in Computational Data Sciences(CDS) at Indian Institute of Science, Bangalore. My research area focuses on Artificial Intelligence at the intersection of Computer Vision, Machine Learning and Deep Learning. I'm currently working at Visual Computing Lab, under the supervision of Prof. Anirban Chakraborty. I am passionate about developing Machine learning/Deep learning models that could solve complex and challenging real-world problems impacting millions of users.

For more about me, check out :

Resume Contact Me

Skills

Languages

Python

C++

Shell Scripting

Libraries

NumPy

Pandas

matplotlib

Frameworks

scikit-learn

PyTorch

Keras

TensorFlow

Other

Git

AWS

Projects

                
Face recognition with limited data
                  A Face recognition model which is trained with limited amount of data(few shot learning).
                
AccomplishmentsLabelled Faces in the Wild(LFW) face dataset was used where
                    some people have a single image and some have multiples. 
Siamese Neural Networks (SNN) was used, where we can compare two images without retrain the network for the new classes.
Incorporated Convolution Neural Networks (CNN) for feature extraction
Tools:TensorFlow, Keras, Kaggle

 Multi class car classification using pretrained CNN models
                  A Multi-class classification model which could classify different car models.
                
Accomplishments Worked on  Stanford Car dataset which has more than 16k images with 196 classes 
After preprocessing of the image data by resizing, used transfer learning approach
Trained the model with few pre-trained CNN models such as AlexNet, ResNet50 and VGG16 
Tools:TensorFlow, Keras, Kaggle

 Efficient stepping strategy for standing bipedal robots under external perturbations 
                  A Efficient trained stepping strategy for standing bipedal robots while external force was applied on it.
                
                Accomplishments
                
                 We used forward and inverse kinematics for getting the desired position of the leg and arm 
 Also used inverse pendulum model for balance the robot
 Used Neuro-Fuzzy system for learning strategy 
Tools:Webots, C, MATLAB