Need Help!

[Jibril Akmel]

hello friends!
i'm jibril, road and transport engineering student!
  I want to analyse image of road pavement distresses, the images look like
below. what i want here in imagej are
1) how can imagej  identify or recognize each distress types ?
2) In general how can i get a java programming that gives me a tool to
analyse and compare pavement road distresses for road maintenance work ?
please i need your urgent response.
respect!

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Aleksejs Fomins]

Dear Jibril,

This is a task for deep learning, not so much for ImageJ, as the
features corresponding to each distress type are highly non-trivial, and
likely too difficult to identify applying standard filters.

Have a look, for example, at tensorflow

https://www.tensorflow.org/tutorials/

It is an easy way to get started with your problem.
1) Take a dataset of images, for which you have already labeled the
distress type by hand.
2) Take a basic deep neural network, and train it using your data, for
example

https://www.tensorflow.org/tutorials/keras/classification

3) See how well it performs on labeled and unlabeled data. Tweak your
deep network accordingly.

Notes:
* Tweaking the network is not easy or intuitive. There is a lot of
experimentation and reading forums involved
* I only recommend Tensorflow because I use it. There are other
comparable alternatives (e.g. pytorch)

Best,
Aleksejs

On 26.11.19 13:14, Jibril Akmel wrote:

> hello friends!
> i'm jibril, road and transport engineering student!
>   I want to analyse image of road pavement distresses, the images look like
> below. what i want here in imagej are
> 1) how can imagej  identify or recognize each distress types ?
> 2) In general how can i get a java programming that gives me a tool to
> analyse and compare pavement road distresses for road maintenance work ?
> please i need your urgent response.
> respect!
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html