Visual and linguistic concepts naturally organize themselves in a hierarchy, where a textual concept "dog" entails all images that contain dogs. Despite being intuitive, current large-scale vision and language models such as CLIP do not explicitly capture such hierarchy. We propose MERU, a contrastive model that yields hyperbolic representations of images and text. Hyperbolic spaces have suitable geometric properties to embed tree-like data, so MERU can better capture the underlying hierarchy in image-text datasets. Our results show that MERU learns a highly interpretable and structured representation space while being competitive with CLIP's performance on standard multi-modal tasks like image classification and image-text retrieval.

When I first applied to Toptal, I wanted to become both a freelancer and a "real ML engineer" at the same time. Before that, I worked as a Machine Learning engineer at Nordeus, a top mobile gaming company famous for having Mourinho's face on its flagship game: TopEleven. My Machine Learning adventure at Nordeus consisted of designing and implementing an intelligent system to help the customer support team resolve player issues faster. The essence of it was to build a text classifier from a ton of historical player tickets and agent resolutions. I had the whole system in mind, the data (at least that is what I thought), and access to GPUs.

Customers often need to identify single objects in images; for example, to identify their company's logo, find a specific industrial or agricultural defect, or locate a specific event, like hurricanes, in satellite scans. In this post, we showcase how to train a custom model to detect a single object using Amazon Rekognition Custom Labels. Amazon Rekognition is a fully managed service that provides computer vision (CV) capabilities for analyzing images and video at scale, using deep learning technology without requiring machine learning (ML) expertise. Amazon Rekognition Custom Labels lets you extend the detection and classification capabilities of the Amazon Rekognition pre-trained APIs by using data to train a custom CV model specific to your business needs. With the latest update to support single object training, Amazon Rekognition Custom Labels now lets you create a custom object detection model with single object classes.

I'm really not ready to go for a long, high speed trip in a completely automated car. I say that because of my baked potatoes. I've done it many times before. Here is my typical process. I take out a variety of vegetables to chop and chop the broccoli, red onion, garlic, red pepper while the potatoes are in the microwave.

Seek is a new app you may want to download if you use an iPhone. It's like a Shazam for nature: the app can help identify the things you photograph using the power of image recognition. The app was created by iNaturalist, an online social network for nature enthusiasts, and its stated goal is to help everyone explore the nature around them. If you ever take a picture and aren't sure of what exactly you just captured, pull out the Seek app and point the Seek Camera at the living thing. Seek will use AI to examine the photo and attempt to match it to one of the roughly 30,000 species it currently recognizes.

Artificial Intelligence (AI) is transforming many human activities ranging from daily chores to highly sophisticated tasks. But unlike many other industries, the higher education sector has yet to be really influenced by AI. Uber has disrupted the taxi sector, Airbnb has disrupted the hotel industry and Amazon disrupted first the bookselling sector, then the whole retail industry. It is only a matter of time then until the higher education sector undergoes a significant transformation. Within a few short years, universities may well have changed beyond all recognition.