Phenome 2019 for Computer Scientists
At the intersection of plant biology and computer science is the phenome, where machine vision, 3D modeling and even entirely new kinds of computational analysis support and advance crop improvement and basic plant science.
Which makes Phenome 2019 the perfect place to discuss and form collaborations between researchers spanning the two fields. In addition to workshops that will lead participants through imaging and autonomous phenotyping, several computer scientists will be sharing the partnerships they’ve developed with plant scientists and the problems they’re addressing together.
We caught up with a couple computer scientists to get a preview of what they’ll be presenting and to hear what they’re looking forward to at the conference.
Tabb says she is looking forward to catching up with the collaborator she began partnering with at Phenome 2018. Those kinds of collaborations are a strength of the multidisciplinary conference, she says.
“The interaction among the attendees (at Phenome) is very good, from the perspective of trying to find collaborators or good problems to work on,” says Tabb, explaining that cross-discipline partnerships take work finding common goals and language. “The payoff can be in establishing interesting new research areas and funding opportunities outside of your previous ones.”
George Kantor, a senior systems scientist at Carnegie Mellon University’s Robotics Institute, will be sharing his work in a talk entitled “Robotics and AI Platforms for Rapid In-Field Phenotyping”. You can watch Kantor speak on agricultural robotics, translational research, and the quest to feed the world’s expanding population in this video from Carnegie Mellon University.
Kantor agrees that Phenome offers new avenues for computer scientists: “I think computer scientists should be excited about the conference because rapid phenotyping is a great domain to pursue computer science research.”
He adds that many phenotyping problems can readily be addressed with existing computational tools. And when you add in interesting problems without known computer science solutions, Kantor says, it makes for satisfying work. “This combination of relatively straightforward adaptation of existing techniques and the need to invent completely new techniques provides lots of opportunities for productive research.”
Ian Stavness of the University of Saskatchewan Department of Computer Science writes “I think it is important for computer scientists to attend this conference because there are many challenging computer science problems that, if solved, have the potential for breakthrough discoveries in plant science. In particular, image-based plant phenotyping from outdoor images presents interesting challenges that push the boundaries of modern deep learning based computer vision algorithms. Compared with other areas, such as image classification with millions of images, we deal with very small datasets of thousands of plant images, which poses a significant challenge for data-hungry deep learning methods. We also deal with outdoor images of plants that have variable lighting due to clouds, cluttered backgrounds due to crop residue on the ground, and plant motion due to wind (it gets very windy in the Canadian prairies!). At Phenome 2019, I'll be talking about the computer tools we have been developing to help capture, organize and make sense of aerial images captured by drones and tractors moving over our plant breeding fields at the University of Saskatchewan.”
“Here is a picture of our Computer Science students getting "dirty" setting up a Camera On A Stick (COASt) for capturing time-lapse images of canola plants growing throughout the summer.”
Want to learn more? Get started at the Phenome 2019 website.
Written by Eric Hamilton, for Peridot Scientific Communications/Plant Editors