With computer vision, a computer can with the help of an operator, perform accurate dimensional analysis of a nematode. The operator can also assist in the computer in identifying esophagus counts and distinguishing features. Each distinguishing feature would reduce the number of total known species to select for a probabilistic match.
Is there such software already developed in Computer Assisted Nematode Identification (CANI)?
I am puzzling obstacles in the biometrics of nematodes and any help is appreciated.
...I know a group working on this in Australia, even they have released a preliminary results, coupe of years ago ...
"The proposed method was tested on 68 images which included 935 live nematodes and 780 dead nematodes and it achieved an accuracy over 85 %." In distinguishing dead and live Heterorhabditis bacteriophora nematodes.
Article A review of recent sensing technologies to detect invertebra...
Nemaplex ( http://nemaplex.ucdavis.edu ) has a key sets for nematode identification.
http://nemaplex.ucdavis.edu/Taxadata/DorylaimidaKey.htm
NEMAID 3 (http://genisys.prd.fr/nemaid) is a computer based methodology facilitating nematode identification.
Most likely in the neighbourhood of about 97.4% accuracy of a nematode that has been classified, photographed and tagged. A more interesting question now would be how well can a computer identify an un-classified nematode -- one without a name?
"Deep learning has solved perception."
François Chollet, author of Keras - keras.io - Keras is a leading deep learning framework for Python, with over 350,000 users and over 700 open-source contributors (as of Q2 2019) and a software engineer at Google:
"What makes deep learning special within the field of machine learning?
In the span of only a few years, deep learning has achieved tremendous breakthroughs across a wide range of tasks that have been historically perceived as extremely difficult for computers, especially in the area of machine perception: extracting useful information from images, videos, sound, and more. Given sufficient training data (in particular, training data appropriately labelled by humans), it’s possible to extract from perceptual data almost anything that a human could extract. Hence, it’s sometimes said that deep learning has solved perception.
...
In that sense, deep learning could be analogous to the internet: it may be over hyped up for a few years, but in the longer term it will still be a major revolution that will transform our economy and our lives.
...
The most surprising thing about deep learning is how simple it is. Ten years ago, no one expected that we would achieve such amazing results on machine-perception problems by using simple parametric models trained with gradient descent. Now, it turns out that all you need is sufficiently large parametric models trained with gradient descent on sufficiently many examples. As Feynman once said about the universe,
“It’s not complicated, it’s just a lot of it."
...
The whole thing hinges on a single core idea: that meaning is derived from the pairwise relationship between things (between words in a language, between pixels in an image, and so on) and that these relationships can be captured by a distance function."
https://www.manning.com/books/deep-learning-with-python
WorMachine: machine learning-based phenotypic analysis tool for worms
Article WorMachine: Machine learning-based phenotypic analysis tool for worms
Computer-assisted identification of nematodes.
Article Computer assisted identification of nematodes
Morphometrical variability in HeZicotyZenchus Steiner, 1945
Article Morphometrical variability in Helicotylenchus Steiner, 1945 ...
@
Results from a k-NN classifier raw pixel intensities of the animals and worms dataset subset. Nematode bodies were imaged with 20x and/or 40x objectives. The entire body was captured.
precision recall f1-score support
Mesocriconema curvatum 0.40 0.34 0.37 29
Mesocriconema inaratum 0.45 0.83 0.59 6
Mesocriconema sp. 0.20 0.30 0.24 10
Mesocriconema xenoplax 0.42 0.38 0.40 13
Ogma sp. 0.17 0.25 0.20 4
Pratylenchus penetrans 0.56 0.83 0.67 6
Xenocriconemella macrodora 0.33 0.20 0.25 15
cats 0.44 0.51 0.47 274
dogs 0.38 0.49 0.43 236
pandas 0.81 0.40 0.54 238
accuracy 0.46 831
macro avg 0.42 0.45 0.41 831
weighted avg 0.52 0.46 0.47 831
Evaluating a k-NN classifier, we see that we obtained 52% accuracy – this accuracy isn’t bad for a classifier that doesn’t do any true “learning” at all, given that the probability of randomly guessing the correct answer is 1/10. There is a lot that can be done to push the accuracy higher. I suspect the Mesocriconema sp. accuracy is being confused by the other two species in the same genus Mesocriconema curvatum and Mesocriconema inaratum.
A well documented approach -- Nematode Identification and Expert System Technology
Edited by Renaud Fortuner
California Department of Food and Agriculture Sacramento,
California and Department of Nematology University of California,
Davis Davis, California
Available as an ebook!
- Badges
- Science topic