Project

Brand logo recognition for real-time social media monitoring

Client

UnamoX

General info

Client

UnamoX helps clients optimize activity across social media.

The SaaS business improves SEO and social media performance via a dedicated analytics suite that monitors user activity across Facebook, Instagram, YouTube, and Twitter.

Project

UnamoX contacted DLabs.AI looking to extend the paid feature set of its social media monitoring tool. 

DLabs created an algorithm that recognizes brand logos in any online image so that UnamoX clients can monitor brand reach and engagement and then contact the right influencers. The system works with 60% accuracy, making it the most effective algorithm of its type.

About the project

Problem

Brands have no way of knowing when their logo appears in images online. Worse, few brands own enough data to train an algorithm to recognize their logo, making building an automated logo detection solution overly expensive.

Solution

An algorithm that can detect a logo in an image (while becoming increasingly accurate as it analyzes more data). And a tool that generates synthetic data to train the algorithm in the first instance.

Results

Together, the synthetic data generator and the logo detection algorithm have become the most accurate logo detection tool on the market today.

Project duration

6 months

What did we do?

See how we prepared data to train neural networks.

See the results

Synthetic data for different logos

  

Technologies used

Python

Django

Docker

Generative Adversarial Networks

TensorFlow

Nginx

OpenCV

PostgreSQL

  

The path to success

Step 1: Generate synthetic data

Develop a tool that could use a logo image to create a synthetic dataset

Use the tool to segment images, perform depth-testing, and carry out surface estimation

Use the tool to apply the logos with the correct size and perspective, placing them in the appropriate segment of the image

Generate tens-of-thousands of images to train the algorithm

Step 2: Train the algorithm

Use deep learning and an array of other technologies to design and train the final algorithm

The new tool learned to detect logos thanks to the synthetic dataset

Step 3: Build a verification tool

The end-user required a verification tool to validate when the algorithm correctly identified a logo

The algorithm uses this second, more precise data set to improve the accuracy of its original logo detection

See it in action

  

CLIENTS OPINION

format_quote

They exhibit great knowledge about machine learning.

Kamil Starski, COO, UnamoX

AI SOLUTIONS WE’RE PROUD OF

See other AI projects that have helped our clients achieve their business goals.

Copy link
Powered by Social Snap