Footfall Count App
Tracking and counting of people from video streams have found its’ usage in stores, malls, stadiums, airports, etc. We made an application that tracked and counted the total number of people entering and leaving a particular region (e.g. a store). There are two basic parts involved in this project. First, to detect the object in an input frame and draw a bounding box around it. And the second part is basically assigning an id to each bounding box and then track those bounding boxes along with the frames.
This project aims to track deforestation using remote sensory data (satellite images). We used two different methods for comparison and validated the trends observed. The first method uses basic image processing where we extracted the forest cover area based on threshold color value (HSV spectrum). In the other method, we used Deep learning, specifically a U-net model to extract the forest cover region for respective satellite images.
The recommender system has become mandatory in recent years for any website that lists multiple products and contents. It basically recommends or suggests new content or products based on user interaction and past activity on that website. We built a recommender system for an e-commerce website based on product information and user interaction data provided by the customer. We used both the content similarity approach as well as user similarity approach in solving this problem.
Changes in codes are standard practice in the software industry. The usual method of testing all the changes is time-consuming and becomes difficult for larger organizations where the change volume is huge. To tackle this problem efficiently, we have built this application which predicts the risks associated with all the changes. We have used change-log data to build this application, however keeping in mind that different organizations have different types of data, this application is data agnostic.