AI News, Horses for Sources

China's AI scientists teach a neural net to train itself

So-called self-supervision is an element that can be added to lots of machine learning tasks so that a computer learns with less human help, perhaps someday with none at all.  Scientists at China's Sun Yat-Sen University and Hong Kong Polytechnic University use self-supervision in a new bit of research to help a computer learn the pose of a human figure in a video clip.  Understanding what a person is doing in a picture is its own rich vein of machine learning research, useful for a whole number of things including video surveillance.

So, the Sun Yat-Sen researchers set out to show a neural network can refine its understanding by continually comparing the guesses of multiple networks with one another, ultimately lessening the need for the 'ground truth' afforded by a labeled data set.  China's AI scientists show how their machine learning model refined its 'prediction' of the 3D pose of an actor from an image by adding some self-supervision code to the last part of the neural network.

Notably, Qian is with SenseTime, the Chinese AI startup that sells software for various applications such as facial recognition, and which distributes a machine learning programming framework called 'Parrots.'  In their original paper from 2017, the authors used an annotated data set, the 'MPII Human Pose' data set compiled in 2014 by Mykhaylo Andriluka and colleagues at Germany's Max Planck Institute for Informatics.

'After initialization, we substitute the predicted 2D poses and 3D poses for the 2D and 3D ground-truth to optimize' the model 'in a self-supervised fashion.'  They 'project the 3D coordinate(s)' of the 3D pose 'into the image plane to obtain the projected 2D pose' and then they 'minimize the dissimilarity' between this new 2D pose and the first one they had derived 'as an optimization objective.'

In a sense, the neural network keeps asking if its 3D model of the body is predicting accurately in three dimensions what it thought at the beginning of the process in two dimensions, learning about how 3D and 2D correspond.  There is a lot of now-standard machine learning stuff here: A convolutional neural network, or CNN, allows the system to extract the the 2D stick figure.

diagram of the full neural network set-up for 3D Pose Machines, including a convolutional neural network to extract 2D figure understanding, followed by long a short-term memory network to extract temporal information key to 3D understanding, followed by a final self-supervised comparison between predictions to improve the results.  Then, a long short-term memory, or LSTM, a neural network specialized to retain a memory of sequences of events, is used to extract the continuity of the body from multiple sequential video frames to create the 3D model.

Meituan Drives Instant Food Delivery With AI “Super Brain”

Instant food delivery — food arriving in 60 minutes or less — is an integration of online e-commerce transactions and offline logistics delivery into a system, with connecting relationships between customers, merchants, delivery drivers, and the platforms themselves.

Abundant reliable data helps to improve the AI system so that it can best predict order arrival times and make better decisions regarding order dispatching, pricing, logistics network design, etc.

Meituan drivers report after each delivery, helping Super Brain collect additional highly dense and reliable location data to augment the information provided by map companies, customers and merchants.

The Meituan AI logistics team has made significant progress over the past three years: average delivery time has been shortened from one hour to 30 minutes, while its intelligent dispatch system path calculation capability has climbed to a rate of 2.9 billion times per hour.

With the urban food delivery market showing no sign of slowing down, Meituan plans to expand Super Brain’s data collection sources while using AI to improve its updating and information integration to deliver an even better experience to diners.

Three Kingdoms AI is Looking Dreadful! And Buggy to Boot!

How do you even create a bug that sees a dude flung from his horse and EXPLODE! when he hits a wall! Sources ...

BINA48 Robot Talks to Evie the Avatar

BINA48, a sentient robot who can see and hear, talks to Evie, an avatar using Cleverbot artificial intelligence. Connect with Ars Technica: Visit ...

A.I. Learns Nobel Prize Experiment in Just 1 Hour!

When A.I. meets science, interesting things occur. The next few years may be very interesting for the scientific community. Subscribe here: ..


Red Dead Redemption 2 NEW DETAILS! Euphoria Engine Physics and AI are vastly improved in Red Dead Redemption 2 with unique interactions with NPCs ...

Introducing WildCat

WildCat is a four-legged robot being developed to run fast on all types of terrain. So far WildCat has run at about 19 mph on flat terrain using bounding and ...

Did Campos Say Peephole or People? - Artificial Intelligence Perspective

What did Jesus Campos say to the Las Vegas Metro Cop as he pointed down the Hallway at the room of Paddock? Did Jesus Campos say Peephole? Or did he ...

Future Military Robots Explained

Warfare is going through its most significant change in human history. This is an in-depth look at how robotics is increasingly preventing soldiers from rich ...

Bina48 Robot Talks to Siri

The BINA48 robot, created by Martine Rothblatt and Hanson Robotics, has a conversation with Siri. Read the article: ...

Humans Need Not Apply

Support Grey making videos: ## Robots, Etc: Terex Port automation: ..

The Rise of the Machines – Why Automation is Different this Time

Automation in the Information Age is different. Books we used for this video: The Rise of the Robots: The Second Machine Age: ..