CVPPP 2014 Home

Computer Vision Problems in Plant Phenotyping (CVPPP)

 

Zurich, September 12, 2014, in conjunction with ECCV 2014 (http://eccv2014.org/)

 

Proceedings available as Springer book LNCS 8928.


The goal of this workshop is to showcase computer vision challenges in plant phenotyping. Plant phenotyping is the identification of effects on the phenotype (ie., the plant appearance and behavior) as a result of genotype differences (ie., differences in the genetic code) and the environmental conditions a plant has been exposed to. Knowing the plant phenotypes is a key ingredient for knowledge-based bioeconomy and this, not only literally helps in the efforts to feed the world, but is also essential for feed, fibre and fuel production.

 

While previously, collection of phenotypic traits was manual, currently noninvasive, imaging-based methods are increasingly utilized in plant phenotyping and resulting images need to be analyzed in a high throughput, robust, accurate, and reliable manner. The occurring problems differ from usual tasks addressed by the computer vision community due to the requirements posed by this application scenario.

Plants are complex, self-changing systems with complexity increasing over time. Typical problems in measuring their actual properties comprise measuring size, shape, 3d surface structure, architecture, and other structural traits of plants and their organs (leaves, fruits, roots etc.) or plant populations, where core problems are e.g. reliable detection and multi-label segmentation of many similar objects, or reconstruction of specular, almost featureless, discontinuous surfaces. When interested in changes over time, for example growth rates, tracking, optical flow or scene flow estimation are required. Inherently the tracked objects change their appearance over time. In some cases images may be acquired under controlled conditions, but typically are taken in challenging environments occurring in the field, in green houses, employing automated and/or affordable acquisition setups.

Unfortunately, without automated and accurate computer vision to extract the phenotypes, a bottleneck is encountered, hampering our understanding of plant biology.

 

Target audience are

  • computer vision experts interested in novel application fields, well accessible to computer vision, but different in requirements
  • plant phenotyping scientists with rich expertise in image processing and computer vision interested in standardization, as exact problem formulations in fact allow defining standards.

download flyer

Phidas logo

 

EU and FP7 Logos

 

DPPN Logo

 

Some example images from data set

IPPN-Logo