These stripes, called Blaschko lines, are invisible most of the time, but certain skin (and other) conditions bring them out. In fact, everyone’s stripes follow the exact same general patterns.
HUMANS WITH STRIPES SERIAL
The author is a serial entrepreneur, techie and promoter of WildTrails, a wildlife travel company.*All* humans have stripes.
HUMANS WITH STRIPES SERIES
And as technology advances, it may soon be possible tell, not just one tiger from another, but also the age and gender of the lord of the jungle, just from a series of pixels. As more images from tiger lovers are fed in, more images from government sources and tiger conservation groups are analyzed, it will be possible to create signatures for every one of the 3,000 and more tigers that grace our forests. A Tiger ID app, or web application will become better as more images are fed into the system. Building a base 96% accuracy may sound impressive but it is nowhere near enough. The model identified Maya correctly at an accuracy rate of more than 96%. This signature ID was used to test the algorithm, again, with thousands of images of Maya, other tigers and other cats. The result was the creation of a signature id that was exclusive to Maya. The algorithm was fed hundreds of images of faces of Maya and other tigers.
The test case was Maya, a popular and well-loved tigress from Tadoba, the popular tiger reserve in Maharashtra. Once the computer could, with 98% accuracy, pick out tigers from other cats, it had to be taught to distinguish between individuals within the species. This not only involved images of tigers, but an equal number of images of leopards, house cats and lions – for not only must the computer learn what a tiger image looks like, but it must also learn what a tiger image does not look like, despite similarities. Species, then individual The next step was to collect a significant number of images for the computer to ‘practice’ the recognition process. And the content of the signature files - stripes on the face, the marks on the forehead, the shape & structure of the face, the eyes, its cheeks, its mouth - also had to be decided on a trial-and-error basis. But for a tiger, these signatures had to be created from scratch. For humans, there a signature files that tell a computer that a certain pattern of pixels means that that portion of the image is an eye, or a nose, or a cheek. But for such a program, you would need to feed the computer signature IDs of tigers – the result of analysis of many images of the big cats, to provide a base of information that the machine can use for identification. A program to do this would simplify the process. But it’s a tedious process, with many steps and prone to error. But what if you want to help a computer distinguish one tiger from another? Back to basics The 2018 tiger census uses the pattern of a tiger’s stripes to tell one cat from another. These signature IDs are created by the analysis of thousands of similar faces and can be reused by programmers who create facial recognition software. This body of work – these so-called “signatures” are available in the public domain.
There is a lot of work that has gone in to help computers identify faces. Training a computer to distinguish one face from another requires the analysis of a different set of parameters than those for training it to recognize the make and model of a car. A computer looks for patterns of pixel values in set of images fed as “training set” of inputs. Today machines can recognize only those features that they are trained for. Classification depends on not just the features the computer needs to consider but also to features which it needs to ignore. And image recognition, much like a human recognition, is all about classification based on a core set of features. Teaching a machine The technical term is machine learning, but the key is how the computer is taught to do something. And this is at the core of image recognition. Recognition is based on the identification of a certain set of common features- as well as disregarding other features as extraneous.
0 kommentar(er)
