New Czech plant identification software claims to be best in the world
A joint effort by two Czech universities claims to have developed the most accurate plant image recognition system in the world. Able to identify thousands of different kinds of plants and mushrooms, the software has already won three international competitions, beating human experts in the process.
The series of competitions at which the software gained its accolades was organized by the annual Computer Vision and Pattern Recognition (CVPR) Conference which took place in Salt Lake City this June. In the various contests computer systems were pinned both against each other as well as against human experts.
The software has the potential to become popular in a country where mushroom picking is a regular pastime. In the summer of 2017, an app based around a similar concept, which claimed to be able to identify 210 species of mushroom with an accuracy of around 60 percent, became the second most downloaded application in the Czech Republic.
“We have already met with several groups. At the conference for example there was a Chinese app for recognizing plants and then we are also in contact with other international teams, some Danish organization focused on fungi recognition, etc. We are also in discussion with more groups, or more teams. I would say that so far we have not reached any specific details yet because these contacts are very fresh.”
Scientists rather than businessmen, Šulc and the cybernetics team are more interested in improving the software, rather than any developing the app.
“For us the priority is still to improve the state of the art in machine learning and computer vision. Maintaining the application is not a priority for us. We would therefore prefer a collaboration with someone interested in running the app and we would provide the technology and artificial intelligence to do this with high precision.”
Using the software’s methodology, the scientists at the Czech Technical University are already working on other projects that may make our lives easier.
“Technologies that are similar to those we use in plant recognition are applied also to other projects. We are for example also working on a project for smart appliances for homes and we can see similarities in the methods between the different applications.”