Blog

Ready to learn how to train your AI in less than a day without writing a single line of code? If you’re sitting on a pile of data that you want to make useful, but AI seems too daunting of a task to get started, then this article is for you.

Or if you have experienced the trenches of AI and are ready to make the model training process a whole lot simpler and faster, you’ll also find this article eye-opening.

Off-the-shelf algorithms have gotten so good that they’ve ushered in a new paradigm of accessibility in machine learning.

Continue reading “AI is More Accessible Than You Know”

In this article, we’re going to discuss Swift 5 value and reference types and see how they’re handled.

Value types are faster to work with since they operate out of the stacks, and copying a value is cheap since it happens in a constant time. Values also help us achieve predictable behaviour and isolation, whereas reference types give shared access to memory locations and dynamic storage.

Continue reading “Swift 5: Value Vs Reference types”

NSLinguisticTagger, which was available as far back as the iOS 5 SDK, paved the path for Apple’s announcement of their Natural Language framework at WWDC 18. Everything from language identification to lemmatization and part-of-speech tagging, all of which were present in NSLinguisticTagger, are now a part of the Natural Language framework, with an API that’s been completely redesigned in Swift.

The added benefit that the Natural Language framework has over the NSLinguisticTagger is the ability to use custom NLP models.

Continue reading “Exploring Word Embeddings and Text Catalogs with Apple’s Natural Language Framework in iOS”

Following up on my blog post on training a TensorFlow Lite model with AutoML Vision Edge, this blog post aims to teach you how to load that .tflite model into an Android app and run inference (i.e. predictions) on it.

If you haven’t read that previous blog on training your own TensorFlow Lite model, I highly suggest that you do so before proceeding with this blog:

Continue reading “Loading and running a quantized TensorFlow Lite model on Android”

Core ML is an interesting means for adding a pre-trained machine learning model to your app. But one question that nagged me after trying Core ML was—How can I possibly train my own model and integrate it in my apps using Core ML?

Well, after doing some homework, a learned a lot about the possibilities of achieving this. To be honest, all the methods require you to understand and know your math really well! While I was on this roller coaster ride, I came across Custom Vision.

What a relief for developers looking to jump straight into training their own machine learning models. With the help of Custom Vision, developers can easily manifest their machine learning ideas into real mobile apps without diving too deeply into the machine learning waters.

Continue reading “Using Core ML and Custom Vision to Build a Real-Time Hand Sign Detector in iOS”

The wave of AI and machine learning is happening just as the dominance of mobile is becoming set in stone. As mobile devices become more ubiquitous and powerful, a lot of the machine learning tasks we think of as requiring months of high-powered compute time will be able to happen right on your phone.

This post will outline why edge devices are increasingly important, and how machine learning works with them.

Continue reading “Machine Learning models on the edge: mobile and IoT”

Hair simulation is a great way to simulate different but realistic hairstyles. The Hair Simulation effect lets you know which hairstyle suits your face, and it’s quite a lot of fun. We can simulate short, long, and different color hairstyles in Lens Studio.

In Snapchat’s Lens Studio, we will use the Hair Simulation template to try and simulate some beautiful hairstyles.

Continue reading “Add Hair Simulation Effect Using Lens Studio”

As an ML engineer or data scientist, sometimes you inevitably find yourself in a situation where you have hundreds of records for one class label and thousands of records for another class label.

Upon training your model you obtain an accuracy above 90%. You then realize that the model is predicting everything as if it’s in the class with the majority of records.

Excellent examples of this are fraud detection problems and churn prediction problems, where the majority of the records are in the negative class. What do you do in such a scenario? That will be the focus of this post.

Continue reading “Dealing with Imbalanced Data in Machine Learning”

There was no shortage of excellent computer vision research presented during the recently-concluded 2020 CVPR conference. In this article, we’ll take a focused look at a couple of those that touched or have implications for mobile or edge-related tasks.

While not all of these papers connect directly to mobile-first applications, their implications for mobile ML are significant. They push forward ML tasks commonly performed on mobile and edge devices, so their advancement is crucial in pushing the industry forward.

Continue reading “CVPR 2020: Research with Mobile ML Implications”