Finding a picture in an image without marking it up?

We often see pictures in images: comics, for example, combine several pictures into one. And if you have an entertainment app where people post memes, like in our iFunny, you’re going to run into that all the time. Neural networks are already capable of finding animals, people, or other objects, but what if we need to find but another image in the image? Let’s take a closer look at our algorithm so that you can test it with a notebook in Google Colaboratory and even implement it in your project.

Over the past 4 years, we’ve developed an adaptation process enjoyed even by ultra-experienced newcomers. Today, we’ll share its main stages and a basic checklist. You can easily adjust them to fit engineers of any level that you hire.

Most articles about the algorithms of the multi-armed bandit are too academic. They are filled with formulas and seem to imply that we have an immutable set of handles for pulling and n→∞ attempts. I will try to talk about these algorithms from the point of view of an ordinary developer, taking into account the real conditions in which our iFunny application for recommending memes works.

We share the zombie object detection mechanism we made for the iFunny app, and present tips for those who also want to get rid of this issue. So you turn on your laptop, open Crashlytics, and voila: EXC_BAD_ACCESS objc_release… What now?

Imagine how much memory space a 10k-pixel comic book takes up. And now imagine that you can’t compress it because if you did, it would lose too much quality and become unreadable. Curious how we at FUNCORP deal with this? Then read on!

We’ve been long working on improving the user experience in UGC products with machine learning. Here are our ten key lessons of implementing recommendation systems in business to build a really good product.

The iOS team of iFunny has gone from a completely ad-free model to using a variety of ad networks and formats in their popular entertainment app. In this article, we will discuss some of the less-apparent nuances of working with advertising SDKs that can affect the user experience and performance of your product and share the code that will help you fix them.

Overview of self-supervised methods.

While the demand for neural networks is growing, most state-of-the-art approaches to adapt them to business needs often lag, hindered by insufficient or absent markup. Supervised learning is hardly feasible in this situation, and standard unsupervised methods won’t work for most of your tasks. This is where self-supervised plans come to the rescue. Depending on the task, they require next to no markup or none at all.

A developer’s perspective (MLOps inside)

At iFunny, we are trying to compose the best possible feed of memes and funny videos. To rate our job, people use smile/dislike buttons and comments. Some of them even post memes about our efforts.

Recommendation systems will always stay relevant — users want to see personalized content, the best of the catalog (in the case of our iFunny app — trending memes and jokes). Our team is testing dozens of hypotheses on how a smart feed can improve user experience. This article will tell you how we implemented the second-ranking level of the model above the collaborative one: what difficulties we encountered, and how they affected the metrics.

Founded on years of work experience in testing

There is a tremendous variety of QA Engineer vacancies, ranging from junior to lead tester and even to principal QA Engineer. We’re often asked what qualities a senior-level tester should have compared to junior or middle-level ones. Let’s try to answer this.

The articles that will come in handy

Approximately 100,000 units of varying content come through our iFunny app daily, and every single one of them needs to be checked. We have already dealt with forbidden imagery by creating a classifier that automatically bans it. Next up — old memes, reuploads, and straight-up doubles that users try to sneak past the moderation.

To get rid of those, we have introduced a duplicate detection system. It had already gone through several iterations, but at some point, we realized it was impossible to put version-to-version improvements in proper perspective. And so we ventured into the Net, searching for books and articles that would allow us to examine currently existing approaches to duplicate detection and — most importantly — to their quality assessment. You can see what we’ve found below.

FUNCORP’s services and advertising infrastructure have recently undergone significant changes. In addition to the Prebid Mobile, we now also support and develop the Prebid Server to work with our apps.

We chose Prebid because we’ve been using its Software Development Kit for a long time and have enough experience and competence in working with it. This expertise allows us to improve and develop the service without compromising its stability. As a result, Prebid has become one of the critical services within our infrastructure.

It is pretty standard that as a project grows, so does the complexity of its build. Too many different technologies, third-party components, libraries, lints, server-side rendering, and project-specific nuances — as a result of all this, the configuration of a build may involve more than a thousand strings.

This article explains how you can automatically divide a dataset of images into clusters classified by qualitative contextual feature, thanks to embeddings from the much talked about neural network called CLIP created by Elon Musk’s company. I will give you an example using the content from our iFunny app.

The right to be forgotten by the User’s request

There are different ways to remove a user’s personal information at the user’s request to make a product compliant with the CCPA or GDPR. The most basic method is to handle every request received by mail manually. The important thing is to make sure the process is as straightforward as it can be and clear to the user. So a little automation is not such a bad idea.

In 2020, for the first time since World War II, the world was faced with an epic, uncontrollable crisis. The overwhelming majority of the world’s commercial industries, including the tech industry, have had to deal with the negative impact of the pandemic.

Apple set the entire mobile advertising market on edge last year when it announced the pending enactment of its new privacy policy for the App Store. From now on, app developers are required to seek users’ permission to track their devices for advertisers (yes, that’s the controversial IDFA). The new deal, delayed several times, was finally enacted at the beginning of this year.

iFunny users upload about 1,000,000 pieces of content to the app every day, including not only memes but also racism, violence, pornography, and other inappropriate material.

Previously, we checked all this manually, but now we are developing automatic moderation based on convolutional neural networks. We have already trained the system to divide content into three classes: it recognizes what can be included in user feeds, what needs to be removed, and what is hidden from the shared feed. To make the algorithms more accurate, we decided to add a specification for removing content that did not have such markup before.

Testing ad integrations is quite a tedious process. Since we work with third-party SDKs that can hardly be controlled, it can be challenging to automate the testing process. Nevertheless, to reduce testing time, we have developed and implemented a debug panel, which we will discuss in this article.