SLIDE 1
Chris
IMAGE RECOGNITION WITH SYNTHETIC DATA FOR RETAIL THERE WILL BE - - PowerPoint PPT Presentation
IMAGE RECOGNITION WITH SYNTHETIC DATA FOR RETAIL THERE WILL BE - - PowerPoint PPT Presentation
IMAGE RECOGNITION WITH SYNTHETIC DATA FOR RETAIL THERE WILL BE NO MATHEMATICAL FORMULAS Chris
SLIDE 2
SLIDE 3
Наличие товара на полке
Наличие товара на полке
SLIDE 4
Наличие товара на полке
Solving business problems
Наличие товара на полке
Correct results Control over the progress Optimizing the work of field staff Reliability of collected data
SLIDE 5
Chris
- Image recognition for objects on a supermarket shelf
is the first and key stage of automation in retail
- We, humans, get 85% of our information through
visual sensors (eyes)
- Computer vision, i.e., training mathematical models to see and
understand what they see, is a fundamental problem of automation
IMAGE RECOGNITION IN RETAIL BY ECR RESEARCH:
ABOUT 40 BLN IMAGES PER YEAR
SLIDE 6
Chris
We give the business
1. Cost reduction 2. Scalability 3. Sales growth
IN OUR SOLUTIONS, IMAGE REGOCNITION CAN BE DONE
ON AN OFFLINE MOBILE DEVICE
SLIDE 7
Наличие товара на полке
CHALLENGES FOR
IMAGE RECOGNITION
IN RETAIL APPLICATIONS
SLIDE 8
Chris
- To automatically recognize goods on the
shelf, we need to design and train deep neural networks
- To teach the model to distinguish
between Pepsi Cola and Coca-Cola, it needs to “see” thousands of labeled images from both classes
- Central problem in many applications of
neural networks: where will the labeled data come from?
LOTS OF LABELED DATA REQUIRED
SLIDE 9
- Current catalogue of goods only in
Russian-speaking (ex-USSR) retail:
- Each SKU requires
- Moreover, the network needs “real-looking” photos,
i.e., labeled products on photos from real shelves...
170 000 SKU 1000 - 5000
labeled photos
WHERE WILL ALL THIS LABELED DATA COME FROM?
SLIDE 10
Chris
- Where can you get 1 billion
labeled photos?
- Today they are labeled by hand, usually
through crowdsourcing services
- A “turker” labels ~50 photos per 8 hour
workday, ~$0.2 per photo
PARADOX: TO AUTOMATE HUMAN LABOR, WE NEED ORDERS OF MAGNITUDE MORE HUMAN LABOR
SLIDE 11
Chris
- Manual labeling – years of work, hundreds of million dollars
- Moreover, crowdsourcing labeling is unreliable (the human factor),
and you need to cross-check, labeling each photo several times
- Wrong data will certainly lead to wrong models
1 BLN LABELED PHOTOS = 120 MLN WORK HOURS
SLIDE 12
Наличие товара на полке
THE NEUROMATION SOLUTION
SLIDE 13
- We create a virtual copy of the shelf (a 3D model) and place 3D objects
- Then we can generate unlimited amounts of labeled synthetic data
- We train deep neural networks on synthetic data
- Reaching good accuracy much faster and cheaper than the competition
SLIDE 14
Advantages of synthetic data
- We generate 100% accurate labeled
data, with pixel-perfect labeling which is impossible to do by hand
- Increase the speed of automation by
- rders of magnitude
- The resulting solution is several
times cheaper than hand labeling
SLIDE 15
Examples
SLIDE 16
Examples
SLIDE 17
SLIDE 18
SLIDE 19
Наличие товара на полке
THE SYNERGY BIG DATA + IMAGE RECOGNITION
FOR FMCG RETAIL BUSINESS
SLIDE 20
Chris
BIG DATA + IMAGE RECOGNITION
OSA Hybrid Platform Image Recognition by Neuromation
Together we develop a new approach to solving business problems for retail through automation
SLIDE 21
Chris
SUMMA TECHNOLOGIA = SALES GROWTH
SLIDE 22
Chris
В НОВОЙ ЭКОНОМИКЕ ДАННЫЕ – ЭТО НОВАЯ НЕФТЬ. ГЛУБОКОЕ ОБУЧЕНИЕ НА СИНТЕТИЧЕСКИХ ДАННЫХ ПОДОБНО ПРОИЗВОДСТВУ СИНТЕТИЧЕСКОЙ НЕФТИ.
IN THE NEW ECONOMY
DATA IS THE NEW OIL
DEEP LEARNING ON
SYNTHETIC DATA
IS A CHEAP AND ABUNDANT
SYNTHETIC OIL
SLIDE 23