Timeline for Neural network: Line detection in Image
Current License: CC BY-SA 3.0
14 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jun 1, 2020 at 14:47 | answer | added | MaxJ | timeline score: 2 | |
| Jun 6, 2019 at 7:25 | answer | added | Yuval | timeline score: 4 | |
| May 11, 2017 at 13:46 | comment | added | MaxJ | Thank you all for you help! Together with the segmentation example I could find a net that performs extremely well. | |
| May 10, 2017 at 21:35 | history | tweeted | twitter.com/StackMma/status/862420828964745217 | ||
| Apr 19, 2017 at 0:37 | comment | added | Yss | The example I was referring to is in the nettrain documentation. I'm not sure of the exact page away from my computer, I think it's applications->computer vision->semantic segmentation. Also, as an alternative to elimianting noise, you can conversely try to add more to get a very robust and noise immune network. In general, if you can see it, a convnet can be made to see it. | |
| Apr 18, 2017 at 21:32 | comment | added | Gregory Klopper | In documentation for NetTrain under Applications> Computer Vision. | |
| Apr 18, 2017 at 21:27 | comment | added | Gregory Klopper | Pre-processing your image even if you do go with a neural network solution, is a must. Training your network in such a noisy image would require a much more robust network and much longer training. | |
| Apr 18, 2017 at 14:14 | comment | added | MaxJ | @Yss Do you have a link to the example? I can not find it. | |
| Apr 17, 2017 at 23:11 | comment | added | Yss | This is a semantic segmentation type problem. The desired output of the network is a heat map (or threshholded heatmap) of whether or not a given pixel belongs to one of your desired pixels. There is an example in the 11.1 docs on semantic segmentation that is a pretty good place to start | |
| Apr 17, 2017 at 19:56 | comment | added | MaxJ | @GregoryKlopper These approaches work very well to denoise the image but as you guessed right I do not want vertical lines. Additionally, the area to the left is quite noisy and it is hard to discern clear lines by eye. Thus I did not trace these by hand and I wanted to train the network to not do that either. | |
| Apr 17, 2017 at 19:52 | comment | added | MaxJ | @DavidG.Stork The lines are not necessary straight but can become quite curved. Also I only want clear diagonal lines and discard noisy areas/lines. Thats why I initially thought the network approach might be best. As far as I understood the Hough Transform is mostly useful to extract broken lines or? | |
| Apr 17, 2017 at 19:44 | comment | added | Gregory Klopper |
@user3683367 Why are you only paying attention to the diagonal lines? Is that a requirement? You seem to be throwing away vertical lines (from noise) which have about same luminosity as the diagonal ones. So how do you know what's noise and what's the real deal? Run these on your source gif and see what you get: Binarize@ImageAdjust[ImageConvolve[ImageAdjust[ImageSubtract[gif,.15],1.05], {{-1,0,1},{-2,0,2},{-1,0,1}}],1.1] or ImageAdjust@GradientFilter[ImageAdjust[ImageSubtract[gif,.1],1.2],{2,.5}]
|
|
| Apr 17, 2017 at 17:56 | comment | added | David G. Stork | Why not use techniques from computer vision such as Hough transform? | |
| Apr 17, 2017 at 17:46 | history | asked | MaxJ | CC BY-SA 3.0 |