deepsight logo


Image Recognition SDK 

Powerful State of the Art image processing through a simplified REST API powered by Deep Learning

Robust Face Detection

Our face detectors are built using state of the art Convolutional Neural Networks that offer high accuracy and next to nil false positives.  Tried and tested in crowded, indoor and outdoor environments.

Facial Landmarks

Extract 68 point facial landmark points such as tips of eyes, nose and lips.  These can be used to perform accurate pose estimation and face alignment.


Extract demographic details using neural networks trained to classify Gender and Age from faces.  Analyze and infer key insights from these metrics about your business and cater to your audience accordingly.

Face Recognition

Our neural network trained on a million faces outputs face embeddings for each face that can be used to identify and recognize individuals.  Extraction of 128 dimensional embeddings takes a few milliseconds per face and can be performed  in realtime.



State of the Art

Deep Learning is state of the art in AI research and has produced excellent results in the field of Computer Vision.  DeepSight leverages this technology to give you powerful image processing capabilities through a simplified interface.

100% Offline

DeepSight runs completely offline as a http service on localhost:5000.  All inference using CNNs is performed on the machine itself.  This allows you to build solutions that can be deployed on remote machines running independently 24×7.


With its simplified RESTful API, building powerful solutions is fast, easy and intuitive

Modular Design

DeepSight was made modular.  Each CNN is a separate entity that is plugged into the main application.  This allows us to provide updated modules with improved accuracy over time without disturbing your existing solution.  It also enables DeepSight to be extended to fit your evolving business requirements.

CUDA Accelerated

DeepSight can use a CUDA enabled GPU to significantly boost performance by up to 10x !

One Click Installation

We provide DeepSight in an extremely easy to install Setup.exe/ package.  No technical expertise is required to install and maintain it.


In Action



No GPU Acceleration
Face Detection
68 Landmarks
Face Recognition
CUDA Accelerated SDK
Per Node
GPU (CUDA) Acceleration
Face Detection
68 Landmarks
Face Recognition
4 CNN Face detectors
Unlimited API Calls
14 Day Free Trial
Contact Us
Training on Custom Dataset
Object detection
Semantic Segmentation
Platform Specific Optimization
End-to-End Solution

Introducing FaceDB

FaceDB is a binary that runs as a http service.  It is capable of storing embeddings in the millions and allows for searching the closest matching embedding within a second.

What does it mean?

It means that you can use FaceDB to query a face from a database of a Million faces and expect results within a second.  This allows you to build highy scalable surveillance applications ranging from banking, industrial, public spaces to commercial sectors.

We’ve benchmarked this with 14M faces (embeddings) on a desktop grade CPU and it searches accurately within 1s.

Number of Embeddings
Search Time
1000 2 ms
10,000 14 ms
100,000 200 ms
1 Million 970 ms