Sean Meighan
Software => Nutcracker Models => Topic started by: keithsw1111 on April 28, 2016, 06:25:15 AM
-
So here is a video update showing where I am at with auto generation of custom models ... this is my best case ... so dont get too excited yet. The snowflake model I was also sent I cannot get a good result from yet ... but it is also much more challenging ... daylight + a white frame. I havent given up yet though.
https://youtu.be/Np6f19A0HnE (https://youtu.be/Np6f19A0HnE)
-
Very impressive sir!
James
Sent from my SM-G900V using Tapatalk
-
Fantastic job for just a few days of working on this new effect. I'm sure given a bit more time you'll get it sorted, I'm amazed at how you and the other developers come up with some of the stuff you do.
Stuart
-
Next you just need it to interface directly to the camera.
-
That is excellent work. That will make it so easy to bring in a custom singing face or snowflake.
I saw someone post an element over on the LOR FB page that is an excellent candidate for such a mapping program. It was a tree with random bulb positions.
One thing I would request is that in the element viewing window that you would be able to crop or zoom in on just the element being examined. Don't know how difficult that might be but you did show that the sides could be dragged in from two sides to eliminate some of the surrounding clutter.
Thanks a whole lot.
-
very nice work!!
-
another suggestion... can you display the pixel count somewhere? it might help detect those double-pixel situations. if it detects 103 pixels and know you only put 100 on the snowflake, you would then look for the duplicates.
-
Next you just need it to interface directly to the camera.
Yeah. Can ffmpeg do that?
-
another suggestion... can you display the pixel count somewhere? it might help detect those double-pixel situations. if it detects 103 pixels and know you only put 100 on the snowflake, you would then look for the duplicates.
That's a good idea.
-
Next you just need it to interface directly to the camera.
Yeah. Can ffmpeg do that?
I don't think ffmpeg is built for that purpose. I did a search and I'd probably start here if you were really interested:
http://larryo.org/work/information/wxopencv/index.html (http://larryo.org/work/information/wxopencv/index.html)
-
One thing I would request is that in the element viewing window that you would be able to crop or zoom in on just the element being examined. Don't know how difficult that might be but you did show that the sides could be dragged in from two sides to eliminate some of the surrounding clutter.
This is more complicated than it would seem. I am also not keen to overly complicate this process. This is about getting a good approximation to the model ... not necessarily a perfect one. The user can always edit it later.
But there are some things that do need to be right:
- I need all nodes ... if i miss nodes then it is hard to add them later as all higher nodes need to be renumbered. Missing a number is ok though as the user can easily add that in (if they can work out which one is missing ... i need to make that easy).
- I need to generate a reasonable sized model. It is hard for the user to shrink or expand the model later.
Doubling up bulbs or misplacing them is not a big deal as that is pretty easy for a user to change themselves.
-
I don't think ffmpeg is built for that purpose. I did a search and I'd probably start here if you were really interested:
http://larryo.org/work/information/wxopencv/index.html (http://larryo.org/work/information/wxopencv/index.html)
I have been trying hard not to use opencv ... I have already added enough libraries to xLights. OpenCV would have saved me from implementing myself a bunch of algorithms ... although it did let me optimise them for our use case.
-
Wow! Very impressive. Will be such a time saver ;D