Plankton community ecology and ecosystem research is currently greatly hampered by the bottleneck of acquiring species-level information from the communities due to the slow analysis process with traditional microscopy methods. The FASTVISION consortium will combine front-line plankton imaging instrumentation and taxonomic expertise of the Finnish Environment Institute and the computer vision and image analysis excellence of the LUT University. The project will use digitized microscopy images of plankton provided by novel automated imaging instruments, to train image recognition software for species identification using efficient machine learning approaches that produce interoperable data across instruments and habitats. This big data of plankton community composition with an unprecedented high resolution will be applied to experimental and field tests of key hypotheses in plankton community ecology, biodiversity and ecosystem functioning.
The FASTVISION project is funded by Academy of Finland. It is a collaborative project between the LUT University and Finnish Environment Institute.
People
Tuomas Eerola
 |
Office: 2425 | Tel: +358 50 404 2054 | Associate Professor, Project Manager |
| Heikki Kälviäinen
|
Office: 2422 | Tel: +358 40 586 7552 | Professor, Project Leader |
| Lasse Lensu
|
Office: 2420 | Tel: +358 40 759 1720 | Professor |
| Heikki Haario
|
Office: 2412 | Tel: +358 400 814 092 | Professor |
Publications
Plankton Recognition in Images with Varying Size
In the paper, various modifications to the baseline convolutional neural networks are compared to address the extreme size variation in plankton image data.
Towards operational phytoplankton recognition with automated high-throughput imaging and compact convolutional neural networks
In the paper, steps towards the operational use of automated phytoplankton classification methods are taken by evaluating and analysing compact convolutional neural networks on plankton image data collected from the Baltic sea.
Download: DOI
Classification of Varying-Size Plankton Images with Convolutional Neural Network
In the master's thesis, the challenge of extreme variation in plankton image size and aspect ratio is addressed by proposing various modifications to the baseline convolutional neural networks.
Download: link
Semi-supervised learning for plankton image classification
In the master's thesis, utilization of semi-supervised learning techniques are considered for plankton recognition to employ large volumes of unannotated plankton image data.
Image clustering for unsupervised analysis of plankton data
In the master's thesis, plankton image clustering is considered with the aim to provide unsupervised analysis of plankton image data.
Download:
Plankton recognition from imaging flow cytometer data using convolutional neural networks
In the master's thesis, a deep convolutional neural network based classification method is evaluated on plankton images collected using imaging flow cytometer.
Download: URN