Researchers developed a new machine learning technique with the potential to improve robotic vision.
A new study conducted by Universitat Autonoma de Barcelona and Carnegie Mellon University researchers reveals a technique that could push deep learning algorithms to learn image features without any annotations by humans.
In order to accomplish this remarkable feat, the algorithms need to be trained using large-scale annotated datasets that harness extensive information on each image. The technique is much more efficient than manually annotating images using human labor.
“We aim to give computers the capability to read and understand textual information in any type of image in the real-world,” said Dimosthenis Karatzas, co-author of the study.
The researchers created the new technique as an alternative to unsupervised algorithms that harness non-visual elements along with images, essentially creating a method of self-supervised training.
“This turns to be a very efficient way to learn how to represent images in a computer, without requiring any explicit annotations – labels about the content of the images – which take a lot of time and manual effort to generate,” Karatzas said. “These new image representations, learnt in a self-supervised way, are discriminatory enough to be used in a range of typical computer vision tasks, such as image classification and object detection.”
“We will continue our work on the joint-embedding of textual and visual information, looking for novel ways to perform semantic retrieval by tapping on noisy information available in the Web and Social Media,” he added.
Karatzas and his team hopes to continue identifying the best methods for harnessing image-embedded textual information to describe and answer questions regarding image content automatically.
The findings were published on the pre-print server arXiv.
‘Humans are underrated,’ says Tesla’s CEO Elon Musk, proceeding the company’s failed attempt to attain weekly production targets in first quarter of 2018.
CEO and product architect Elon Musk publicly explains one of the reasons towards Tesla’s unsuccessful first quarter of 2018. Musk states that automation has been hindering Tesla’s Model 3 production due to a stronger focus on robotics rather than humans, resulting in “excessive automation”.
During a tour around Tesla’s factory, CBS presented inquiries concerning a correlation of robots and slower production instead of heightening manufacturing speed. Musk replied, “Yes, they did. We had this crazy, complex network of conveyor belts and it was not working, so we got rid of that whole thing.”
Additionally, Musk explained, “Yes, excessive automation at Tesla was a mistake. To be precise, my mistake. Humans are underrated.”
As a result of production failure towards 2,500 Model 3 vehicles, in the first quarter of 2018, uncertainty escalates for Tesla’s ability to reach its 5,000-a-week target in three months time. In return, the shortfall has delayed crucial customer deliveries. Musk described his forced actions to take direct control of the production line at the beginning of April, with recourse to working through the night and sleeping at the factory.
In the meantime, Tesla is confronting adverse public relations involving a fatal accident of one of its Model X SUVs that was operating using the company’s Autopilot mode.
The herringbone structure enables the Mantis shrimp to deliver a punch similar to a 22-caliber bullet.
The Mantis shrimp is a beautiful sea creature that would not be found in aquariums due to its ability to shatter glass. Suffice its small size, and the crustacean packs a massive punch, able to kill much bigger creatures. But in theory, an animal measuring less than 20 centimeters should not be able to shatter glass or kill even small animals in the sea with just a punch.
However, the mantis shrimp is an apex predator to this respect. Its blow can be seen sending water repels and heard for quite a distance underwater. Scientists are now looking at how this majestic sea creature can achieve such vicious blows.
Head researcher Professor David Kisailus has attributed this ability to a calcium based structure called herringbone structure. It is made from calcium and phosphorus that has crystallized forming a light yet extremely hard structure. This component is also found in the human bone though in lower concentrations. The herringbone structure enables the Mantis shrimp to deliver a punch similar to a 22 caliber bullet.
“This unique herringbone structure not only protects the club from failure,” said Professor David Kisailus. But also enables the mantis shrimp to inflict incredible damage to its prey by transferring more momentum upon impact.”
The researcher, implemented his data to a 3D projection to see how exactly they can use the information in mechanics. He believes that the first and most straightforward implementation would be construction helmets. The technology would enable the protective gear to be stronger and lighter, making the workers safer and more comfortable.