Start your AI journey with us - Deep Learning Analytics

All Services

Computer Vision

We are experts at building state of the art solutions using the latest in AI. We have experience with deploying many types of computer vision models for clients in a variety of verticals. Contact us to learn more:

Expertise with:

Image Classification – Classifying images into different categories is a fundamental task in computer vision and deep learning models are getting to human level accuracy in classification. Read about how we classified road signs into different classes for self driving cars here.
Face Detection and Recognition – Through advances in facial detection research, machines are now able to detect human faces in images. Many businesses are starting to use facial recognition as a biometric security, understand emotions from faces and much more. Read about how we built a pipeline for real time face detection, identification and emotion recognition here.
Object Detection – With deep learning we can now detect and localize objects in images and videos using architectures like SSD, Faster RCNN, YOLO, RetinaNet etc. We have built over 50 custom object detection models for different applications in Retail, Sports, Health care, manufacturing etc. Read about how we combined object detection with opencv based analytics for better understanding of soccer games here.
Aerial detection – With the cost of drones decreasing, there is a surge in aerial data being generated. Aerial object detection is a an important yet challenging problem in computer vision. We used a RetinaNet to build a powerful aerial pedestrian detection model. Read more here.
Object Tracking- Object tracking involves tracking an object across a series of frames as they move across the scene. Object tracking can be a starting point to enable more detailed understanding of user behavior, example if a store tracks how its customers move around and shop. Read about how we used the deep sort algorithm to allow for simultaneous tracking of multiple pedestrians here.
Object Segmentation – Object Segmentation implies understanding pixel wise location of an object in an image. This allows for a deeper understanding that is needed for many tasks. Read about how we built a model for understanding the extent of damage in a car here.
Audio based classification – Audio data is everywhere and is being used for many machine learning applications. This is so that our models can not only “see” their surroundings but “hear” them as well. We are also increasingly using audio to interact with smart agents like Siri and Alexa. We used CNNs and RNNs to build a powerful audio classification model. Read more here.
Handwriting recognition – We can use AI to now read printed and handwritten texts on forms and documents allowing us to automatically “scan” and save this information to a database. Handwriting can also be used for recognizing writers. We built a tensorflow model for this. Find out more in our blog here.
2D/3D pose – We have used deep learning to map 2D and 3D pose for humans playing sports allowing for better understanding of player techniques.

Natural Language Processing

Language is around us everywhere and increasingly businesses are using AI to build smart applications that understand text. We have experience with deploying many types of NLP models for clients in a variety of verticals. Contact us to learn more:

Expertise with:

Text Classification – With the recent advances in deep learning, the ability of algorithms to analyse text has improved considerably. We used a CNN to understand customer complaints and built a smart case router to direct them to an appropriate department. We gave a talk about this at Google Next 2018. Listen to it here.
Text Summarization – Text Summarization is a powerful capability to summarize the information around us from news, blogs, emails etc. Read more how deep learning sequence to sequence models enable text summarization in our blog here.
Question Answer Modeling – How amazing would be a model that can answer questions from any paragraph by extracting word(s) from the paragraph that are most relevant. Deep learning has made this possible! This has applications everywhere. See our model in action on our demo’s page. Read about how this is done at our blog here.
Topic Modelling – Topic Modelling is the task of using unsupervised learning to extract the main topics that occur in a collection of documents. This allows for better understanding of main themes in a document. We used Latent Dirichlet Allocation (LDA) for fast topic modelling. More discussed in this blog.
Sentiment Analysis – Sentiment analysis is increasingly being used by businesses to understand the social sentiment associated with their brand, product or services and monitor online conversations. We have implemented sentiment analysis for understanding customer conversations with call center agents.

Structured Machine Learning

Use advance computer vision techniques to recognise and track objects.
perform OCR, analyse videos and much more

Expertise with:

Big Data Analytics – Deep Neural Networks outshine traditional techniques in handling data that is big in both number of data points and large in terms of number of features. Neural Networks have been used successfully for mining big data and uncovering useful trends.
Time Series Modelling – Deep learning methods offer a lot of promise for time series forecasting, such as the automatic learning of temporal dependence and the automatic handling of temporal structures like trends and seasonality. We have used time series data to build classification and regression models.
Anomaly Detection – Anomaly detection (or outlier detection) is the identification of rare items, events or observations which raise suspicions by differing significantly from the majority of the data. We have experience in using variational autoencoders for unsupervised learning of anomalous trends.