Atharva Bankar

Showroom Management - Database Management System

Detection of people and weapons from video footage or images. The detection time was 30ms - 35ms per frame with Darknet-YOLOv4. A custom dataset was used for the training of the object detection model and data integration was performed using OIDv4 ToolKit.

Autoencoder & Variational Autoencoder for data augmentation and checking data authenticity with ML models.

Implementation of SOLO (Segmenting Objects by Locations) Instance Segmentation algorithm from scratch

App Dev & Amazon Sagemaker ML model deployment - Alternate Crop Recommender according to Kharif/Rabi season

Object Detection & Image Segmentation on Satellite Images

Implementation of YOLO (You Only Look Once) Object Detection algorithm from scratch

Implementation of VAEs and GANs from scratch.

Implementation of Cycle-GANs from scratch for paired image translation.

Implementation of transformer based SegFormer-B0 from scratch for semantic segmentation.

Implementation of Faster-RCNN Object Detection algorithm from scratch

Adversarial attacks on CNNs using gradients of the network

Speaker Independent Automatic Speech Recognition for continuous audio.

Machine Learning models for helping BNP Paribas Cardif accelerate its claims process

Uptane: Secure Framework for Automotive Software Updates: Reference Implementation and Demonstration code

Using NLP or prediction of stack overflow posts using linear models for multi-class classification