Image Database for Object Recognition

Link

ImageNet provides a large image database organized according to the WordNet hierarchy in which each node of the hierarchy is depicted by hundreds and thousands of images. Currently an average of over five hundred images per node are available. You can download preprocessed features, bounding boxes or attributes even the images are available these are for noncommercial use only though.

Tweet about this on TwitterShare on Google+Share on FacebookPin on PinterestShare on RedditShare on LinkedInShare on StumbleUponEmail this to someone

The Bionic Kangaroo

The German automation specialist Festo and its Bionic Learning Network have spawned another creature: The Bionic Kangaroo. After recreating the motion mechanics of several fascinating aquatic and airborne animals they seem to now move on trying to learn the tricks nature applies to land based creatures.

With the BionicKangaroo, Festo has reproduced the unique way a kangaroo moves. Like its natural model, it can recover the energy when jumping, store it and efficiently use it for the next jump. kangaroo03

It combines pneumatic and electrical drive technology to produce a highly dynamic system. The stable jump kinematics plus the precise control technology ensure stability when jumping and landing. The consistent lightweight construction facilitates the unique jumping behavior.

The system is controlled by gestures with a Myo Armband. Festo paid particular attention to the mobile energy supply on the artificial kangaroo. For this purpose, the team even developed two different concepts – one with an integrated compressor and one with a mobile high-pressure storage device.

kangaroo04

There is a detailed brochure on the project here. Altogether as always great marketing leading up to the Hannover Messe. Still impressive technology though.

Tweet about this on TwitterShare on Google+Share on FacebookPin on PinterestShare on RedditShare on LinkedInShare on StumbleUponEmail this to someone

An Apelike Balancing and Walking Robot

Researchers at the University of Bremen, Germany have developed “iStruct” an apelike robot that is able to balance and walk on 4 as well as 2 feet.

istruct

The aim of the project is to develop a robotic system as well as biologically inspired structural components which, if applied to a robotic system, effectively improve its locomotion and mobility characteristics. All elements are designed and build as self-contained as possible decentralizing sensing, control and communication. The robot itself is supposed to be a test platform for foot and spine structures.

Balancing

Walking

Transition from four to two feet

Tweet about this on TwitterShare on Google+Share on FacebookPin on PinterestShare on RedditShare on LinkedInShare on StumbleUponEmail this to someone

A Flying Jellyfish

This contraption, built by applied mathematicians Leif Ristroph and Stephen Childress of New York University, is not the first small ornithopter — a flying machine capable of hovering by a flapping-wing motion, such as that of dragonflies and hummingbirds. What distinguishes Ristroph and Childress’s craft from other flapping insectoid biomimikry robots is that it can remain stable in flight using the movement of its wings alone, without the need for additional stabilizers or complex feedback control loops to avoid flipping over.

Tweet about this on TwitterShare on Google+Share on FacebookPin on PinterestShare on RedditShare on LinkedInShare on StumbleUponEmail this to someone

Gesture Recognition without a Camera

allsee

Researchers at the University of Washington have developed AllSee the prototype of a hand gesture recognition system based on measuring changes in the ubiquitous electromagnetic field generated by wifi, tv stations, mobile phones etc. This allows for a low power gesture recognition solution that may even operate when integrated in a cell phone that you’ve put in your pocket.

Tweet about this on TwitterShare on Google+Share on FacebookPin on PinterestShare on RedditShare on LinkedInShare on StumbleUponEmail this to someone

Artificial Muscles made from Fishing Line

An international team led by The University of Texas at Dallas has made the astonishing discovery that ordinary fishing line and sewing thread can be cheaply converted to powerful artificial muscles. This might possibly be an interesting alternative to memory metall (NiTi) based micro actuators.

The new muscles can lift 100 times more weight and generate 100 times higher mechanical power than a human muscle of the same length and weight. They can generate astonishing 7.1 horsepower per kilogram.

In a paper published in Science, the researchers explain that the powerful muscles are produced by twisting and coiling high-strength polymer fishing line and sewing thread. Scientists at UT Dallas’ Alan G. MacDiarmid NanoTech Institute teamed with scientists from universities in Australia, South Korea, Canada, Turkey and China to accomplish the advances.

The muscles are powered thermally by temperature changes, which can be produced electrically, by the absorption of light or by the chemical reaction of fuels.

 

Silverplated Nylon using the silver as heating element and water for rapid passive cooling.

Tweet about this on TwitterShare on Google+Share on FacebookPin on PinterestShare on RedditShare on LinkedInShare on StumbleUponEmail this to someone