Microsoft’s new vision-language model outranks humans at image captioning

The company will integrate the new model into Azure Cognitive Services to support vision-language tasks

Microsoft sign on a building

Microsoft researchers have developed a new object-attribute detection model for image encoding: VinVL (visual features in vision-language)

Vision-language (VL) systems make it possible to search relevant images for a text query (or vice versa). They also help describe an image’s content. 

In most cases, the systems use two modules to achieve the VL understanding: an image encoding module to generate feature maps of an input image and a vision-language fusion module to map the encoded image and text into vectors in the same semantic space.

Microsoft’s new research focuses on improving the image-encoding module. When combined with VL fusion modules such as OSCAR and VIVO,  Microsoft’s newest VL system scored big on the most competitive artificial intelligence (AI) benchmarks, including visual question answering (VQA), Microsoft COCO Image Captioning, and novel object captioning (nocaps). 

The tech giant also highlighted that VinVL significantly surpasses human performance on the nocaps leaderboard for consensus-based image description evaluation (CIDEr). 

Microsoft trained its VinVL object-attribute detection model using a large object detection dataset containing 2.49 million images ascribed to 1,848 object classes and 524 attribute classes to achieve the results mentioned above. Microsoft formed the dataset by merging four public object detection datasets (COCO, Open Images, Objects365, and VG).

Related Resource

Getting started with Azure Red Hat OpenShift

A developer’s guide to improving application building and deployment capabilities

How to improve application building and deployment capabilitiesDownload now

“We first pretrained an object detection model on the merged dataset, and then fine-tuned the model with an additional attribute branch on VG, making it capable of detecting both objects and attributes,” said Microsoft.

“Our object-attribute detection model can detect 1,594 object classes and 524 visual attributes. As a result, the model can detect and encode nearly all the semantically meaningful regions in an input image, according to our experiments.”

Despite the promising results, Microsoft said its model is by no means close to the human-level VL understanding. 

Microsoft also announced VinVL would be available to the public for general use. Additionally, it will integrate VinVL into Azure Cognitive Services to power a wide range of Microsoft services, including Image Captioning in Office and LinkedIn, and Seeing AI.

Featured Resources

Shining light on new 'cool' cloud technologies and their drawbacks

IONOS Cloud Up! Summit, Cloud Technology Session with Russell Barley

Watch now

Build mobile and web apps faster

Three proven tips to accelerate modern app development

Free download

Reduce the carbon footprint of IT operations up to 88%

A carbon reduction opportunity

Free Download

Comparing serverless and server-based technologies

Determining the total cost of ownership

Free download

Recommended

The true story behind the IBM Personal Computer
Hardware

The true story behind the IBM Personal Computer

3 Dec 2021
Microsoft hit with formal complaint over "monopolistic" software bundling
collaboration

Microsoft hit with formal complaint over "monopolistic" software bundling

29 Nov 2021
Gmail vs Outlook.com: Which one is better?
email providers

Gmail vs Outlook.com: Which one is better?

26 Nov 2021
Business customers can get 30% off the Surface Laptop Go for Black Friday 2021
Laptops

Business customers can get 30% off the Surface Laptop Go for Black Friday 2021

26 Nov 2021

Most Popular

What should you really be asking about your remote access software?
Sponsored

What should you really be asking about your remote access software?

17 Nov 2021
How to move Microsoft's Windows 11 from a hard drive to an SSD
Microsoft Windows

How to move Microsoft's Windows 11 from a hard drive to an SSD

24 Nov 2021
Sabbath hackers are targeting US schools and hospitals
ransomware

Sabbath hackers are targeting US schools and hospitals

29 Nov 2021