Added Edge Detection module

[tech4GT]

@jywarren please have a look , right now exporting the edge detected image, will create a separate issue for adding support for json output, what say?

[jywarren]

Oh, this is cool! Could you try running it on the image here and writing back in that thread? I'd like to get some input from these folks:

https://publiclab.org/questions/jlev/02-01-2018/how-can-i-identify-bits-of-plastic-from-the-beach-in-an-image

This is so cool! @ccpandhare what do you think?

[tech4GT]

@jywarren there is one limitation though, this only works with png images right now, but i will add support

[tech4GT]

@jywarren right now this is basic edge detection we can use algorithms like hysteresis to improve the quality of edge detection😁

[tech4GT]

@jywarren since jpg does not support transparency , can i convert an input jpg file to png and convert the output back to jpeg, will that be an acceptable solution?

[ccpandhare]

This is so cool!!! @jywarren One more off the bucket list. :-)

…

On Sat, Feb 10, 2018, 03:17 Varun Gupta ***@***.***> wrote: @jywarren <https://github.com/jywarren> since jpg does not support transparency , can i convert an input jpg file to png and convert the output back to jpeg, will that be an acceptable solution? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#168 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AT0xnL91o0muRqlQ17SS7I9lCIO_JG8sks5tTLzWgaJpZM4SAXrx> .

[ccpandhare]

Why isn't this working with JPEG? On Sat, Feb 10, 2018, 08:30 Chinmay Pandhare <pandharechinmay882@gmail.com> wrote:

…

This is so cool!!! @jywarren One more off the bucket list. :-) On Sat, Feb 10, 2018, 03:17 Varun Gupta ***@***.***> wrote: > @jywarren <https://github.com/jywarren> since jpg does not support > transparency , can i convert an input jpg file to png and convert the > output back to jpeg, will that be an acceptable solution? > > — > You are receiving this because you were mentioned. > > > Reply to this email directly, view it on GitHub > <#168 (comment)>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/AT0xnL91o0muRqlQ17SS7I9lCIO_JG8sks5tTLzWgaJpZM4SAXrx> > . >

[jywarren]

Oh this is cool! I echo chinmay's question! Please post to the plastics question, they'll be interested to see this!!! On Feb 9, 2018 10:03 PM, "Chinmay Pandhare" <notifications@github.com> wrote:

…

Why isn't this working with JPEG? On Sat, Feb 10, 2018, 08:30 Chinmay Pandhare ***@***.*** > wrote: > This is so cool!!! > > @jywarren One more off the bucket list. :-) > > On Sat, Feb 10, 2018, 03:17 Varun Gupta ***@***.***> wrote: > >> @jywarren <https://github.com/jywarren> since jpg does not support >> transparency , can i convert an input jpg file to png and convert the >> output back to jpeg, will that be an acceptable solution? >> >> — >> You are receiving this because you were mentioned. >> >> >> Reply to this email directly, view it on GitHub >> <#168# issuecomment-364578404>, >> or mute the thread >> <https://github.com/notifications/unsubscribe-auth/ AT0xnL91o0muRqlQ17SS7I9lCIO_JG8sks5tTLzWgaJpZM4SAXrx> >> . >> > — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#168 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AABfJ1my4MdkjnQ2-qpKdVU3fZd64lP8ks5tTQbegaJpZM4SAXrx> .

[tech4GT]

@jywarren @ccpandhare actually jpegs do not support transparency values and edge detection requires that
a values are 255 for each pixel in jpeg
I apologize for late reply

[tech4GT]

@jywarren should i post the complete code on the plastics question?

[jywarren]

Cool! I think you could just post the image of edges and link back here.

…

On Feb 10, 2018 7:14 AM, "Varun Gupta" ***@***.***> wrote: @jywarren <https://github.com/jywarren> should i post the complete code on the plastics question? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#168 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AABfJ2rQxtYsg43hFkUvrJY4Vro5VT1wks5tTYgQgaJpZM4SAXrx> .

[tech4GT]

@jywarren done😁

[jywarren]

Ok, maybe we need to specify exactly what the output of the module is. It's a greyscale image?I guess I'm more accustomed to seeing edges as black pixels surrounding white regions but this is cool.

[tech4GT]

@jywarren right now i have just implemented Gaussian blur with sobel filter on the image after conversion to greyscale hence basic edge detection, i will implement remaining steps for canny edge detection in future which will finally output image with black edges surrounding white regions

[tech4GT]

@jywarren right now this outputs a greyscale image you are right😁

[jywarren]

That's awesome. Im thinking some of this explanation could go into the module description to help people understand what it does and what the output means. What do you think?

…

On Feb 10, 2018 10:21 AM, "Varun Gupta" ***@***.***> wrote: @jywarren <https://github.com/jywarren> right now this outputs a greyscale image you are right😁 — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#168 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AABfJ2GcjToe4cgA83HIMeao3lvqJrQWks5tTbO3gaJpZM4SAXrx> .

[tech4GT]

@jywarren sure no problem but right now I don't have access to a computer I'll do it a bit later if that's fine?😅

[jywarren]

Oh of course. Me neither, I'm on my phone :-)

…

On Feb 10, 2018 10:40 AM, "Varun Gupta" ***@***.***> wrote: @jywarren <https://github.com/jywarren> sure no problem but right now I don't have access to a computer I'll do it a bit later if that's fine?😅 — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#168 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AABfJ66tN4KQejjLa-46ybR15fyxSfgxks5tTbhQgaJpZM4SAXrx> .

[tech4GT]

@jywarren please have a look if anything else needs to be added, rest assured this is ready to be merged and i will surely open issues for enhancing this module and adding tests, thanks a lot😁

[tech4GT]

@jywarren is there anything else i can do in this😅

[ccpandhare]

Apart from this small thing, this looks fine to me!
What do you say, @jywarren?

[ccpandhare]

Please fix the double "a" in the description field's value

[ccpandhare]

Again, I apologize for being that picky!

[tech4GT]

@ccpandhare no apologies needed in fact thank you for being so observant😁

[ccpandhare]

Also, how would the edges thing work? Would it be like thresholding the greyscale output image of this module or would it have an entirely different approach?

[tech4GT]

@ccpandhare i was thinking once i implement complete canny edge detection we can get the edge pixels and output the positions, what say?

[jywarren]

Hi, this is looking really good; i wasn't previously familiar with how edge detection works, so I had to do some research to understand fully.

With that in mind, I think we might add this to the description just for clarity:

Edge detection: this module detects edges using the Canny method, which first Gaussian blurs the image to reduce noise (amount of blur configurable in settings as `options.blur`), then applies a number of steps to highlight edges, resulting in a greyscale image where the brighter the pixel, the stronger the detected edge. Read more at: https://en.wikipedia.org/wiki/Canny_edge_detector

Do we also allow setting the threshold? https://en.wikipedia.org/wiki/Canny_edge_detector#Parameters

Finally, the Wikipedia page shows a black background instead of alpha. Would it be possible to do this too? Then it'd work in JPG too.

This is really impressive work! Potentially we could base a Blur module off of components of this too!

[tech4GT]

@jywarren thanks a lot and yes we already have everything for the Gaussian blur so we can create a module for that.
secondly i am yet to implement all the steps of canny edge detection and hysteresis is one of them which takes the threshold values, hence i called this basic edge detection and i will create separate issue for adding the rest of the steps and i will shortly add them all
Thirdly the black background thing can be implemented after the rest of the steps are implemented which i am already working on and i will submit the PR in a week or so, i hope that's fine.

[tech4GT]

@jywarren i have also taken care of the changes that @ccpandhare mentioned😁

[tech4GT]

@jywarren finally done with this, this was a tough one😅. Full canny edge detection🎉

[tech4GT]

[tech4GT]

@jywarren now this also takes low and high threshold and we also have a edge pixel positions array which can be exported as json as well

[jywarren]

Hi, this sounds great and looks great. If you can make a follow-up issue to make the background black, and for any other follow-up issues, then this sounds good and I'm merging it in! Thanks!

[jywarren]

I just wanted to say that this was a very impressive project! 🎉 🚀 👍

[tech4GT]

@jywarren thanks a lot😁