Aug 2020 – May 2021 · 9 mos · Fort Collins, Colorado, United States

Jun 2020 – Aug 2020 · 2 mos · Greeley, Colorado, United States

Jan 2020 – May 2020 · 4 mos · Fort Collins, Colorado Area

• Working on building models to convert low resolution geostationary satellite images to high resolution using Generative Adversarial Networks.

May 2018 – Jul 2018 · 2 mos · Varanasi Area, India

• I was the leader for a group of four. We are given the task of modeling a system for identifying Vehicles ahead of us in real time. We looked at different methods available to achieve this task and found out that YOLO is the best possible method that suits our problem based on our hardware and available datasets. Here in YOLO, we frame object detection as a regression problem to spatially separated bounding boxes and associated class probabilities. A single neural network predicts bounding boxes and class probabilities directly from full images in a go contrasting to the RCNN's.. Since the whole detection pipeline is a single network, it can be optimized end-to-end directly on detection performance. Even though, YOLO makes more localization errors but is less likely to predict false positives on background which is very essential in real time. Finally, YOLO learns very general representations of objects. It outperforms other detection methods. As our YOLO model is trained in COCO classes it can identify around 80 classes. We, customized it to vehicles and it can be applied to both images and videos. We also added an additional feature to the model. Apart from just identification it warns the user when he is going too close to the vehicle or if a vehicle is approaching him and is very close to him.