Stuck in the middle of ADS model process #775
-
|
Hi, I have installed the AxonDeepSeg you developed on my Macbook Pro and my iOS system is macOS Monterey (ver. 12.7.1). All the best, 
[TEM images.zip](https://github.com/axondeepseg/axondeepseg/files/13516531/TEM.images.zip)
|
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 2 replies
-
|
The TEM images are in the zip file. It's strange that I can not attach those Tiff files individually so I compress them to zip file. |
Beta Was this translation helpful? Give feedback.
-
|
Hi @Benson-Hsueh - it seems that because your resolution is an order of magnitude different from the resolution of the TEM training images we used, that the software is dividing your image into a lot of resampled "patches" (~224). You can adjust this type of thing with the "Zoom factor" in our settings, I just tried with a Zoom factor of 0.1 and it seems to have segmented your images decently well (you can adjust the options to see if you can do better; Zoom of 0.1 gave me 4 patches to segment, so increasing it slightly might improve things a a bit). As for your other image that you sent, the contrast between myelin and axon (and background) is very low (everything is bright), so either there was a staining issue or microscope issue there, it would be difficult to segment that type of image using the software without some manual image editing on your end to improve the contrast before trying our tool. Hope this helps! |
Beta Was this translation helpful? Give feedback.
-
|
Dear @mathieuboudreau thanks for your reply. Is there any advice for the zoom factor? For example, if we are using 40X or 60X magnification under TEM to acquire the images, how much zoom factor do you suggest us for the analysis? |
Beta Was this translation helpful? Give feedback.
-
|
Hi @Benson-Hsueh. Here, there are some tips on how to use the zoom factor and the sweep to get the best results. As mentioned by @mathieuboudreau, I'm pretty sure you could enhance your segmentations just by enhancing the contrast on your images. This blogpost talks about the basics for image contrast enhancement. |
Beta Was this translation helpful? Give feedback.
Hi @Benson-Hsueh - it seems that because your resolution is an order of magnitude different from the resolution of the TEM training images we used, that the software is dividing your image into a lot of resampled "patches" (~224). You can adjust this type of thing with the "Zoom factor" in our settings, I just tried with a Zoom factor of 0.1 and it seems to have segmented your images decently well (you can adjust the options to see if you can do better; Zoom of 0.1 gave me 4 patches to segment, so increasing it slightly might improve things a a bit).
As for your other image that you sent, the contrast between myelin and axon (and background) is very low (everything is bright), so either …