CynLr_Logo

Our content is optimally designed for viewing in landscape mode. Please toggle auto-rotate and rotate your device to view.

We're
Hiring!

Arrow_Down

Problem

CynLr
Cybernetics Laboratory
CynLr_Home_Image

CynLr


Visual
Object
Intelligence


A Machine Vision Platform
Crafting Object Intelligence
For Robotic Arms

Pick

any object


Orient

from all
positions


Place

in every way

Arrow_Down

Why?

Problem

Every Manual Labour is Repetitive

Pick->Orient->Place

Tasks Performed on Unstructured Objects

Yet there is no universal automation solution!

Arrow_Down

How?

Why?

Arrow_Right

Gaps in Current Technology

31%

of time is spent in performing physical tasks by world’s labour force.

In the US alone, this amounts to a spend of

$1.27 Trillion

Tasks

Manual Labour is Most Difficult to Automate!

Today, to produce any material across all sectors, manual labour alone consumes 31% of effort & time - the largest among all type of tasks, despite years of advances in automation. That’s because, only repetitive tasks for handling simple, symmetrical objects have remained technically feasible to automate till date.

While every kind of manual labour can be generalized as a series of repetitive, Pick->Orient->Place actions handling objects, however these objects in the real-world are present in a lot of clutter and randomness. This makes the physical task to be performed unpredictable and the automation of such tasks difficult or impossible.

McKinsey_Report

A McKinsey Report on the State of Automation establishes that technical feasibility of automating unpredictable physical work is below 50% in all sectors, yet comprises a majority of the time spent by humans at work.

Read the full McKinsey Report on State of Automation Here!

Arrow_Left

Manual Labour is Hard to Automate

3D Bin-Picking Can Only Approximately Pick & Drop

Deep Neural Networks Have Dismal Accuracies

3D Bin-Picking solutions available in the market perform basic pattern-matching algorithms on sparse 3D depth-map data. Therefore they work only when object geometries are simple, with no occlusion or entanglement, with atleast one part fully visible in the viewing angle trained for, and can only approximately pick and drop objects.

The purpose of picking an object during a manual task is almost always to place in a desired orientation, hence approximate picks and drops finds very limited use-cases.

View existing 3D Bin-Picking product demos on Youtube to verify

Deep Neural Networks for image analysis have evolved for object identification use-cases, and follow a typical Classification-> Localization-> Detection-> Segmentation approach not suitable for object manipulation applications.

Automation of manual tasks demands 99.999% accuracies and repeatability, which traditional deep learning models cannot achieve even with billions of images of training data. It is practically infeasible to obtain training data for all possible scenarios of even a single object.

Deep-learning implementation for robotic object manipulation is still only a research subject with no commercialized solutions.

Arrow_Down

Team

How?

Arrow_Right

Application Overview

Hardware + Software Solution

CynLr Visual Object Intelligence for Manipulation

CynLr_Platform

CynLr Visual Object Intelligence platform for manipulation is a validated proprietary framework of visual intelligence, purpose built for enabling object manipulation applications.

The CynLr soution integrates off-the-shelf articulate robotic arms of any make with proprietary vision hardware purpose built for object manipulation.

Using a multi-dimensional segregation approach, co-ordinated and adaptive acquisition enabled by an integrated eye-brain platform, and deep-neural network models reinforced with motion feedback, the CynLr platform can accurately Pick and accurately Place even complex geometry objects, even when presented in a random bin with occlusions and entanglements.

Arrow_Left

CynLr Solution

Picking, Orienting and Placing objects that are present in random orientations is a complex cognition problem. Neither 3D depth imaging techniques nor deep learning networks on 2D image data have suffient fidelity to perform the complete task universally for different objects.

CynLr Visual Object Intelligence for Manipulation framework achieves this task in a universal, repeatable and scalable manner, through multi-dimensional visual information construction, and a manipulation framework that feeds-back to and assists visual cognition.

Con_Rod_Bin

Objects with complex geometries, such as this connecting rod in a bin, results in infinite number of unpredictable orientations for picking

Con_Rod_Identify

Because of asymmetry and uneven mass distribution, objects easily get entangled making only a partial face visible for the most ideal part to be picked.

Con_Rod_Pick_Points

This needs the system to understand and predict all possible gripping points on the revealed side, also considering hidden parts and choose the best part for picking.

Con_Rod_Hi_Def

The part must then be selected based on the best picking orientation that can deliver most accurate placement. This requires more visual data than just colour and 3D depth map.

Con_Rod_Pick_Place

Trajectory for gripping, oriented pick and oriented place, to be computed & fed to robot as suited to pick the connecting rod of an untrained random orientation in a bin.

Arrow_Down

Contacts

About

Arrow_Right

We are Hiring!

TEAM

Gokul_NA

N A Gokul, CTO

All things technology, and chief visionary

Nikhil_Ramaswamy

Nikhil R, CEO

All things business, and chief execution

CynLr_0.0

CynLr 0.0

Hard worker and quick learner!

You

You?

We are hiring! Check for openings

INVESTORS

TECHNOLOGY ACCELERATOR

Speciale_Invest Arali_Ventures GrowX_Ventures CIIE_Initiatives Dr._Vijay_Kedia Nvidia_Inception_Program

AWARDS & MEDIA

Yourstory_Tech30 TiE_IoT
Arrow_Left

About Us

We are hiring people with either or a combination of these three core skillsets…

Algorithms

Machine Vision &
Reinforcement Learning

GPU Optimization    

CUDA, cuDNN, Algo Optimizations, Parallel Computation

Software

Design, Development &
Architecture,
C++, Multi-threading

Contacts

CONTACTS

Nikhil Ramaswamy | nikhil@cynlr.com
Gokul N A | nagokul@cynlr.com

INCUBATED AT

Nasscom_CoE_IoT

Diamond District, Lower Ground Floor, DD3, HAL Old Airport Rd, ISRO Colony, Domlur, Bengaluru - 560008