tag:blogger.com,1999:blog-7345806147365425073.post5949776695115902631..comments2024-02-29T07:31:17.654+01:00Comments on Peekaboo: Segmentation Algorithms in scikits-imageAndreas Muellerhttp://www.blogger.com/profile/10177962095394942563noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-7345806147365425073.post-70349470610299984312013-07-02T20:25:20.605+02:002013-07-02T20:25:20.605+02:00Either from their website:
http://ivrg.epfl.ch/sup...Either from their website:<br />http://ivrg.epfl.ch/supplementary_material/RK_SLICSuperpixels/index.html<br />Or use my python wrappers: https://github.com/amueller/slic-pythonAndreas Muellerhttps://www.blogger.com/profile/10177962095394942563noreply@blogger.comtag:blogger.com,1999:blog-7345806147365425073.post-84806865262901072952013-07-02T20:21:18.867+02:002013-07-02T20:21:18.867+02:00Hmm, ok, that's a nice clue. Can you point me ...Hmm, ok, that's a nice clue. Can you point me to the reference code?JNIhttps://www.blogger.com/profile/11763785850061383439noreply@blogger.comtag:blogger.com,1999:blog-7345806147365425073.post-16530371067069412762013-07-02T08:22:19.814+02:002013-07-02T08:22:19.814+02:00Sorry, I didn't know you were working on it. T...Sorry, I didn't know you were working on it. That's great. I have no idea where the difference comes from. I has been on my todo for a long time but so are many other things :-/<br />First I thought it was the way that the Lab conversion is calculated, but I don't think this is the case.<br />It seems somehow to have to do with the scaling of the compactness parameter. In the reference implementation, the compactness parameter is pretty robust wrt image size. In my implementation, it is not :-/<br /><br />I'd be glad to help you if you have any idea how ;)Andreas Muellerhttps://www.blogger.com/profile/10177962095394942563noreply@blogger.comtag:blogger.com,1999:blog-7345806147365425073.post-30616253679601742552013-07-02T01:37:06.845+02:002013-07-02T01:37:06.845+02:00Hey Andy,
As you might have noticed I've been...Hey Andy,<br /><br />As you might have noticed I've been messing with your scikit-image SLIC code quite a lot recently. =) I hadn't realised the differences in the superpixel shapes between your implementation and the original. Do you have any ideas about where the differences come from? I'll be happy to keep working on making this a reference implementation of SLIC! =)JNIhttps://www.blogger.com/profile/11763785850061383439noreply@blogger.com