How Perceptly Uses Image Similarity To Fasten Up Annotation

Perceptly helps Ad agencies to analyze the real reason of what and why their creatives worked or not. Its creative optimization engine identifies messaging, visual, and copy elements that are driving performance.

They basically aim at providing personalized report insights from ads. They help in letting their customers know how effective the ads that you are running are with the help of particular tags that are mentioned in the ads. Along with offering campaign planning and recommendations.

The Challenge

The client generates insights from the ads, but different ad copies contain different elements, which are hard to classify. The client has a lot of ad data in the form of images and videos from which they generate insights to create reports or further planning campaigns.

However, the major challenge that they faced was that it was quite difficult to identify 50-60 tags per ad, which was beyond human efficiency. The ad copy contains some elements that need to be classified.

How Labellerr helps them automate the tagging

Labellerr's automated training data generation engine helps the human labelers team to simplify the manual mechanisms involved in the tagging project.

It has various features like

1. Copying the most relevant file's tagging on the current file based on image similarity and other factors. It saves more than 70% of time compared to complete manual tagging.
2. Quick review to verify the annotation done.
3. Inter-annotator agreement to point out the tags that are subjective in nature and use consensus-based tagging.

A variety of charts are available on Labellerr's platform for data analysis. The chart displays outliers in the event that any labels are incorrect, or to distinguish between advertisements.

We accurately extract the most relevant information possible from advertisements, more than 95% of the time, and present the data in an organized style. Having the ability to validate organized data by looking over screens.

Perceptly uses our platform's capability to tag data from multiple scoial media for different clients.

Each ad Copy contained some specific elements which required classification. They created around 70+ questions that have to be labeled on the ad copy. Labellerr helped with object classification from numerous ads that contained more than 50 specific tags in them. We provided them batch-wise data that will help them with obtaining better results in understanding ad performance.

Results

Perceptly performance of analyzing data is enhanced and fastened which will help them in providing better results through their analytical engine.

Object-oriented classification will help them in fastening their process of analysis with our data annotation platform.

In order to guarantee extreme accuracy with the best results and make the computer vision model more trustworthy for end users, a correct training dataset with adequate data is required that we have offered them with monitoring their data using specific tags detected from the elements from ads.