#SciFri: How good is pattern matching software?
We are currently living through some pretty tough times, not since the Second World War has there been the same level of restrictions on movement of people as there are at the moment. As someone that studies disease (in animals), I fully understand the need for the restrictions in order to keep the old and vulnerable safe. Now I’m not here to mention the virus that has been causing global havoc, instead I’m here to distract you and to invite you to take part in one of my research projects. I can guarantee it’s the ideal thing to do if you’re self-isolating and need a moment to take your mind off of things.
You may remember year I posted about my adventures, playing snake with snakes (if not, you can read up on those fun times here). After a while, it got me thinking – just how good are pattern matching software algorithms really? To test this, I’ve been using my friends and guinea pigs and now I’m extending the invitation out to the wider community no matter if you’re a university professor or a business consultant. As long as you’ve got passion for the natural world, I have something to keep you busy for a short while. All I require is an hour of your time to go through and use Wild-ID (the software I’ve been using) to match up 35 pairs of my straightened and cropped photos. I’ve done all the hard work for you, all you need to do is put the kettle on and sit down to play snap!
So how does it work I hear you ask? Well I’ve got a number of different datasets I can send you, each containing 100 photos. The dataset you’re given will be selected by a random number generator to help remove any biases. The 100 photos each contain 35 pairs of snakes and 30 snakes that have no matches. It’s your goal to find all 35 pairs and then send me your results when you’re done. Each of the different datasets has undergone some form of manipulation (except the control group), whether that be a reduced resolution, rotated 90 degrees or transformed to greyscale. I am interested in seeing if any of these manipulations effect the ability of Wild-ID to correctly detect pairs within a dataset. Sound good so far?
But I don’t know how to use Wild-ID I hear you cry! Exactly, that’s what I’m looking for, naive participants to help remove my observer bias. I’ve created a set of instructions that I’ll also send to you that will guide you through everything and if you’re having any issues, please get in touch and we can troubleshoot them together. There is no need to install any additional software as Wild-ID is Java-based and can be run straight from everything I send you (as long as Java is installed). Finally, I’m particularly looking to engage as many people from within the conservation world as possible who may be able to utilise Wild-ID for their own research once they know what they’re doing. This could be in reptiles (like myself) or in any species that has a unique pattern.
If you’ve read all of this and it sounds like something you’re certainly up for, please contact me via Twitter (@stevoallain) or use the ‘Contact’ form. I’ll get to you as soon as I can and provide support where needed. Despite the gargantuan spanner in the works a particular virus has created, we will manage to get through everything. Best of luck matching snakes and stay safe!