RevovleAI

We created an AI-powered Glove detection system. A safety and prevention system that is designed to work for any industry, pandemic situation or business. Wearing gloves is a key safety measure for so many reasons and in so many use cases especially high-risk scenarios like hospitals, food or even construction.
We designed a system that would for any industry by tracking people with an assigned id, checking if they are wearing gloves and generating alerts against a person’s id if they are not wearing gloves.

We created a Face Mask detection system combining the best of Facial recognition and computer vision.

The system tracks people, analyzes faces and generates alerts if necessary. A state of the art Facial recognition system with real-time performance, we implemented it in C++ to guarantee stable and seamless deployability. It can work on multiple cameras simultaneously making it equally ideal for small businesses, public sectors (healthcare, transport, schools etc.) and large enterprises with use cases like smart cities.

We started our journey with a laser focus on what we had to deliver. Starting from training the face detector and facial landmark detector, we wanted to achieve a seamless virtual experience in which a person could try on any product without any delay in augmentation.

We use advanced computer vision and deep learning algorithms to segment parts of the human body. These segmented parts were projected in a virtual space, processed and then cast back to the actual space of the webcam.

The geniuses at RevolveAI worked round the clock to train the State Of The Art ML model on large amounts of data and achieved a fully functional AI system that was fast, reliable and robust.

Caroogle is an AI-powered analytics platform that helps businesses grow by providing intelligent business insights. It helps auto dealers to make quick and efficient decisions based on previous data.

The goal of this project was to determine different voice variations occurring during a speech. These variations can be high, low, pause, soft, stretch, etc. For training an audio classifier, labeled data was necessary, so an audio annotator tool was developed. The annotator tool was able to annotate the audio files using two methods:

• Time-based annotations

• Word-based annotations

In time-based annotation, the user was free to annotate any segment of the audio clip with the pre-selected annotations. For the word-based annotations, the audio was first processed to find the word time-stamps, and then the words and their corresponding time-stamps were shown to the annotator to annotate each word.

After the annotations against each word are acquired from hundreds of audio clips, an audio classifier will be trained to classify each spoken word into one of the sound variations.