If you are interested in looking at magnified pollen

A meeting place for rose breeders.
henry kuska
Posts: 763
Joined: Thu Dec 15, 2011 9:06 pm
Contact:

If you are interested in looking at magnified pollen

Post: # 67591Post henry kuska
Thu Apr 12, 2018 5:55 pm

I hope the actual program is available somewhere. I did not look for it. I assume a good starting point would be to directly contact one of the authors.

See:

https://www.sciencedaily.com/releases/2 ... 102916.htm

https://pubs.acs.org/doi/10.1021/acsphotonics.8b00146

https://www.google.com/search?num=100&n ... pTRoSPsNcQ

Don
Posts: 1732
Joined: Wed Dec 31, 1969 7:00 pm
Contact:

Re: If you are interested in looking at magnified pollen

Post: # 67595Post Don
Fri Apr 13, 2018 2:41 am

> I hope the actual program is available somewhere. I did not look for it. I assume a good starting point would be to directly contact one of the
authors.

The software isn't the half of it. They cite this paper in which they describe using Google's tensorflow AI platform.

https://arxiv.org/ftp/arxiv/papers/1705/1705.04709.pdf

The process involves first developing an AI using parameters tailored to the specific image types, in their case histological slides. They fed both high resolution and then low resolution versions of the same slides into the AI and trained it to recognize the former when presented with the latter, adjusting the various fudge factors along the way. When they got it trained up they then fed it low resolution test slides.

The point is that the AI is easily implemented using tensorflow if they would hand you their python algorithms, but the training is a pia. You would need to feed the AI high resolution images of pollen and then low resolutions version of those same images iteratively and guide the AI in its learning. You would also need a bracket to hold the slides in position on the camera/phone, which is a bit of kit they made using a 3-d printer.

That said, these folks are out of UCLA so if one of them happens to be an amateur rose breeder have them get in touch, my pc looks to be configured capably for the task...
What doesn't kill them makes them stronger.

henry kuska
Posts: 763
Joined: Thu Dec 15, 2011 9:06 pm
Contact:

Re: If you are interested in looking at magnified pollen

Post: # 67598Post henry kuska
Fri Apr 13, 2018 10:49 am

Another article about deep learning (not related to roses but just for general reading):

https://phys.org/news/2018-04-deep-supe ... cells.html

Don
Posts: 1732
Joined: Wed Dec 31, 1969 7:00 pm
Contact:

Re: If you are interested in looking at magnified pollen

Post: # 67632Post Don
Tue Apr 17, 2018 9:57 am

This topic just got a boost. Google has released a do=it-yourself AI toolkit that includes everything you need to train up a visual AI incuding the camera for $89, available through Target.

https://www.target.com/p/-/A-53417081

https://aiyprojects.withgoogle.com/vision
What doesn't kill them makes them stronger.

henry kuska
Posts: 763
Joined: Thu Dec 15, 2011 9:06 pm
Contact:

Re: If you are interested in looking at magnified pollen

Post: # 67648Post henry kuska
Wed Apr 18, 2018 9:52 am

Download the ebook to learn about:
Machine learning vs. deep learning
Convolutional neural networks (CNNs)
Using a pretrained network like AlexNet for image recognition and image classification
Access to examples, tutorials, and software to try deep learning yourself"

See:

https://www.mathworks.com/campaigns/pro ... ous.google

Post Reply

Who is online

Users browsing this forum: No registered users and 4 guests