 |
| Facebook's SEER optimizes image processing for AI computers |
Facebook announced that it has implemented an artificial intelligence program called SEER that can display its search content by submitting more than a billion public photos from Instagram.
Over the past decade, Instagram has become one of the largest image databases in the world, and Facebook, the owner of the company, is using this treasure to transmit the content of automated photos.
According to Facebook, the SEER software outperforms current AI models in object recognition tests.
When performing the test provided by ImageNet, the software scored 84.2% for its classification accuracy. ImageNet is an extensive visual database for finding objects in visual recognition software.
This standard mainly tests whether the AI software can determine the content of an image.
Although several AI models were trained on classified datasets, according to Facebook, SEER learned to recognize objects in images by analyzing random, unclassified, and unsaturated Instagram photos.
Facebook researchers write that this artificial intelligence technology is known as self-learning: The future of AI is to create a system that can learn directly from the information presented without relying on it. Rich datasets to teach them how to recognize images or explain things. A section of text.
They added: SEER performance shows that self-supervised learning can play a prominent role in real-world computer vision tasks. This achievement paves the way for the development of more flexible and accurate computer vision models in the future.
Although this is just a research project, a Facebook spokesperson said: The potential uses are relatively large.
The company stated that potential uses include improved auto-generated text describing images for the visually impaired, better automatic classification of products sold in the Facebook marketplace, and better systems to keep malicious images off the Facebook platform.
We are informing Instagram account holders through a data policy that we are using the information we should be using to support research and innovation, said Priya Goyal, a software developer at Facebook's AI Research Lab.
“While we share research details and build an open source library that allows other researchers to train image models using self-learning, we will not share images or create a SEER,” Goyal added.
Other big tech companies, including Google and Microsoft, are also trying to push the boundaries of computer vision.
Last summer, Google released SimCLRv2 computer vision model and OpenAI iGPT 2.