Anthill Inside 2019

A conference on AI and Deep Learning

Tickets Propose a session

Modeling the effects of blurriness in mobile ads

Submitted by Abhijith C (@abhijith-c) on Tuesday, 30 April 2019

Section: Crisp talk Technical level: Intermediate Session type: Lecture


The creative is the image of the ad that the user sees and engages with upon viewing. This talk studies the effect of an ad creative’s specifications and quality of render in performance campaigns and suggests a playbook for digital marketers based on the findings and insights. An offline study on an ad creative’s specifications such as resolution, aspect ratio, handset density, device orientation, etc. on fit (slot size x creative size) and quality of the render (blurriness / sharpness, etc.) to provide digital marketers with helpful insights.

Further, exploring the state of the art Super Resolution Generative Adversarial Networks for resolving the ad images so that even with an inventory of ad images that do not fit the slot sizes, these ad images can be super resolved and then served to improve on the performance as against the case where the ad images might be blurred out, otherwise.

Most often, an advertiser shares the same set of creatives with multiple ad networks and programmatic channels. These networks and DSPs then bid for the same impressions from a publisher with those same set of creatives. Needless to say, neither ad networks, nor advertisers benefit from this bidding war.

It’s clear from the above example that Media is a commodity; ad impressions can be bought from any media buying channel. The value being added by an ad network or DSP in this model is in matching demand with supply.


  1. Premise/Context
  2. What usually happens when it comes to rendering ad images?
  3. What we found out that is optimal in terms of performance?
  4. Sneak peek into the changes that can be performed to improve performance.
  5. SRGAN: Crash Course
  6. How we super resolve ad images.
  7. Questions

Speaker bio

Abhijith C. is a Data Scientist at InMobi Technologies with a prime area of focus on images. He’s a graduate from Sir. M. Visvesvaraya Institute of Technology, Bangalore (Computer Science). His area of interests are deep learning and computer vision. Abhijith has experience as a Machine Learning intern at Robert Bosch along with a stint at CDS, Indian Institute of Science. He is also an avid competitive coder. Football is his other love, and a Chelsea fanatic since a decade.




  • Abhishek Balaji (@booleanbalaji) Reviewer 5 months ago

    Hi Abhijith,

    Thank you for submitting a proposal. Who is the intended audience for this talk?

    To proceed with evaluation, we need to see detailed slides and a preview video to supplement your proposal. Your slides must cover the following:

    • Problem statement/context, which the audience can relate to and understand. The problem statement has to be a problem (based on this context) that can be generalized for all.
    • What were the tools/options available in the market to solve this problem? How did you evaluate alternatives, and what metrics did you use for the evaluation?
    • Why did you pick the option that you did?
    • Explain how the situation was before the solution you picked/built and how it changed after implementing the solution you picked and built? Show before-after scenario comparisons & metrics.
    • What compromises/trade-offs did you have to make in this process?
    • What are the privacy, security, ethical and regulatory considerations made when designing this?
    • What is the one takeaway that you want participants to go back with at the end of this talk? What is it that participants should learn/be cautious about when solving similar problems?
    • Is the tool free/open-source? If not, what can the audience takeaway from the talk?

    We need to see the updated slides on or before 21 May in order to close the decision on your proposal. If we do not receive an update by 21 May we’ll move the proposal for consideration at a future event.

Login with Twitter or Google to leave a comment