Author: Fritz

In [a previous tutorial], the Levenshtein distance was used in an Android app to suggest words as a user types to automatically complete the word and save time.

In this follow-up tutorial, the suggestions will be enhanced by comparing the text the user is typing to words that are likely to be a match. Moreover, the app will support early exit if the distance exceeds a pre-defined threshold, which helps save time and enhance the suggestions.

Continue reading “Enhancing Word Suggestions for Auto-completing Text in Android”

Document processing is one of the most time- and resource-consuming tasks, compared to the value added that comes from the output.

It takes hours of manual, diligent work to input data in invoices, update medical records, or create the folders for insurance claims.

If all of these tasks could be done automatically, we’d see important benefits, including better organization of the underlying data, automatic scanning of records, error-checking, and pattern detection.

Continue reading “Why AI Is the Next Step in Document Processing”

For a moment, travel back in time with me to 2011. Back then, Kotlin was a garage project by a team at JetBrains and had already been under development for a year. The team’s main goal was to have something that had all the goodness of Scala but still compiled as fast as Java. They open-sourced the project in early 2012 under the Apache 2 License.

Continue reading An Introduction to Kotlin on Android — Concision and Safety

Is your model not performing well? Try digging into your data. Instead of getting marginal improvements in performance by searching for state-of-the-art models, you can drastically improve your model’s accuracy by improving the quality of your data.

Continue reading “It All Boils Down to Training Data”

I’ll be honest with you—I’m horrible at all things art. I was the kid who had to stay after class in 7th grade art because I could not, for the life of me, figure out shading. After years of half-hearted, failed attempts at scratching something decent onto a sketch pad, I resigned myself to the fact that I just wasn’t meant to be an artist.

One of the things I find most fascinating about technology is the ways in which it can clear pathways I would’ve imagined blocked. I’ve found that to be true as I’ve learned more and more about machine learning and mobile development, certainly.

Continue reading “Community Spotlight: Cupixel”

On my not-too-shabby laptop CPU, I can run most common CNN models in (at most) 10–100 milliseconds with libraries like TensorFlow.

In 2019, even a smartphone can run “heavy” CNN models (like ResNet) in less than half a second. So imagine my surprise when I timed my own simple implementation of a convolution layer and found that it took over 2 seconds for a single layer!

It’s no surprise that modern deep learning libraries have production-level, highly-optimized implementations of most operations.

Continue reading “Anatomy of a High-Performance Convolution”

Deep Learning algorithms are excellent at solving very complex problems, including Image Recognition, Object Detection, Language Translation, Speech Recognition, and Synthesis, and include many more applications, such as Generative Models.

However, deep learning is extremely compute intensive—it’s generally only viable through acceleration by powerful general-purpose GPUs, especially from Nvidia.

Continue reading “Building an Image Recognition Model for Mobile using Depthwise Convolutions”