Transform Your Business With Robust Artificial Intelligence Solutions

Developed an AI-powered recommendation engine to provide relevant, personalized suggestions to users based on their unique preferences and browsing behavior, be that for movies, TV shows, series, podcasts, music, books, matrimony, dating, social feed, and more. Businesses of all sizes from different industries leverage this innovative recommendation solution to achieve maximum business value.

For an augmented reality mobile game, team Debut worked on spoken keyword detection with complete implementation on mobile devices using Tensorflow Lite/Keras framework. Requirements included that the model is tiny to be packaged with a mobile app. We implemented a Wavenet model with dilated convolutions, smoothed loss, to achieve over 80% recall with less than 2 false positives per hour of recorded audio.

Implemented a general task-oriented chatbot with custom model architecture, that was able to learn to query and incorporate relation databases to make natural language conversation. Developed a deep learning model that was able to beat state of art models from leading researchers at Stanford University by over 10 percent points on BLEU score on a public dataset by the same researchers.

For an insurance provider, we built additional features for an existing object detection system to identify broken/scratched and dented car parts from car images. Using novel data augmentation techniques and neural net model improvements implemented over a short time frame of a single month, we were able to improve the Average Recall and Average Precision across all damage type categories by 8% and 4% respectively.

Commercial banks have adopted OCR systems to automate the classification and data extraction from the scanned document. Opting for computer vision OCR systems that can read unstructured documents in varying templates and then help in automating the document classification, data extraction and processing would significantly increase the accuracy levels.

Automatic number plate recognition, using the image processing technique of Computer vision indicates new ways of capturing the traffic data. With Automatic Vehicle identification technology and non-stop toll collection methods, a more efficient system can be built. Such systems play an important role in automatic toll collection- by developing systems that can detect text of the license plate from different geometries, sizes, and fonts.

By adopting computer vision in the field of valuable data extraction, physical and social distancing can be monitored. The solution would work by monitoring the physical and social distancing among the public. Based on the data accessed, it would further send alerts in case the protocols are being violated. Confirm the number of people in a given facility and check the congregation limit while maintaining social distancing.

Face masks are one of the best defenses against the spread of COVID-19, but their growing adoption is breaking the facial recognition algorithms. Object Detection and Classification are two major tasks behind this use case. By detecting the facial features of a person crossing the camera and comparing it with the already saved data, AI can enable us to recognize people wearing masks.

Computer vision is also predominantly used in medical imaging such as CT scans, X-rays, and MRI. CT scans are used to detect tumors, internal bleeding, and other life-threatening conditions. A recent study at the University of Central Florida proved that while proficient doctors had only 65% accuracy in detecting lung cancer, the machine was right in 95% of the cases.