Show HN: Image segmentation and background removal in JavaScript(github.com)

60 pointsby secondtimeuse11 anni fa13 comments

Show HN: Image segmentation and background removal in JavaScript

https://github.com/AKSHAYUBHAT/ImageSegmentation

8 comments

This is an alpha version, I hope to improve it over the next week by incorporating ConvnetJS, to label pixels in mixed superpixels (Superpixels with both positive and negative labels). And incorporate additional functionality to such as blurring the boundary.

The current restriction on the size of canvas 640 X 640 (Instagram size) , will also be removed over next few updates.

me_bx11 anni fa

Congrats! Looks beautiful, makes the whole process so easy. This will save so much time to many people.

MichaelApproved11 anni fa

It was hard to get an idea of the final result in the demo gif because the last frame disappeared so quickly. Here it is http://imgur.com/fduYXCi

secondtimeuse11 anni fa

Thanks, I added it to the readme.

skstronghold11 anni fa

Was working on a similar problem, to fetch color of cloths from images(including models and background) in an e-commerce website. This implementation gives the best result so far.

secondtimeuse11 anni fa

Depending on the size of the object you are trying to segment relative to the image, you can adjust the size of superpixels. To segment parts such as clothes you can set large superpixel size by enabling developer option on top right corner. You can also see superpixels rendered to understand which size gives best results. This method can perform even approximate non-interactive segmentation.

MichaelApproved11 anni fa

secondtimeuse11 anni fa

It will be released under Open Source compatible license, I just need to make sure since I am a PhD student.

Note that the underlying algorithm implementations are MIT licensed.

https://github.com/kyamagu/js-segment-annotator/blob/master/...

pmontra11 anni fa

Great demo. Out of curiosity, why not compiling the C algorithm with emcc into asm.js and probably get better performances (but a much larger js file) ?

secondtimeuse11 anni fa

Its actually based on https://github.com/kyamagu/js-segment-annotator which implements two algorithm SLIC and PF.

Regarding compiling C using emscripten, I wanted a pure JS implementation so that it was hackable using Chrome Developer tools, E.g. using LAB data internally produced by SLIC algorithm. Hopefully in future, I will use a emscripten implementation. It might be important for version that proceses short gifs or videos.

Also there is a developer mode on top right, enabling it allows visualization of superpixels and change parameters of the algorithm.

eonw11 anni fa

i tried to use it, it just crashed FireFox 39.0 on Windows 7.

dharma111 anni fa

Great stuff. Anyone aware of Android libs for this?

secondtimeuse11 anni fa

OpenCV has GrabCut implementation which works on android. There are numerous implementations of SLIC superpixels. Even this will work through a cordova style wrapper, but I would really not recommend it, since you can use a much faster and accurate native OpenCV on android.

pg_is_a_butt11 anni fa