|
What you see is really what you get ??
| |
|
 |
 |
|
| |
Hi,
When comparing the vignette to one of the working file ('OUT'.gif) and the contour you mignt be surprised by some shapes that ZooImage extracted. Most images are clearly defined but some of them can bring you in deep water (...) In others words many of the contour are not similar in shape to the vignettes associated with them. This can bring error in the identification into the training set.
We use sample from 202um in which a lot of variation in shape is seen (organics debris, small copepod, large copepod, etc). For example, two particles that seem separated on the vignette are viewed and the working file have one image with one contour. Be carefull when sorting vignette for the training set.
Philippe,
Can we imagine to see the contour ovelapping the vignette to confirm what we see is really what we get ?
Thanks. 
Katherine and Pierre
|
|
|
|
|
|
|
| |
|
|
|
|
| |
Hi there
Would it be possible to show any image examples? It sounds like the mask is somewhat different from what you see on the vignette, but i would have thought that any "particle" that is identified would have its own vignette?
|
|
|
|
|
|
|
| |
|
|
|
|
| |
Hi,
We'd like to show you some images that may be of some concern to those using Zooimage in the identification process. By looking at the vignettes you would assume the contour would be tight around the image but as we went through the _work directory (out.gif) and matched the vignette (particle) with its associated contour we were suprised to see a contour that did not resemble the image as closely as we thought.
The following images (to follow after permission) are from stained samples. Note these examples have been chosen for their extreme contrast(note: most of the contours are correct). We understand that without attention these images will be misclassified because unknowingly ZooImage has seen a larger particle than what we see.
Example 1. A stained sample using lower than 10% ETOH mix for the scan leaked some stain. This method of staining using ETOH mix may be more problematic than just water alone. The contour treats the leaked stain as part of the image creating a contour with little resemblance to the actual shape of the particle.
Example 2. Another instance where we have found contours that don`t exacly resemble the particle seen is when two particles are very close together but visably seperate by a space. There are two ways that ZooImage may interpret this. 1) If the two images are contained in one pixel this creates one contour around both particles. 2) When using stain the background can be tinted causing a bridge between the two particles again resulting in one contour for two particles.
In conclusion Philippe would it be possible to save the out.gif (_work directory) and include the contours in the same directory as the Zip and Zim file to help us in creating more accurate categories and/or overlap the contours with the vignettes.
PS Jen can you allowed us to add the file containing our example. What format we need to save the file ??
Joyeux Noel!
Katherine & Pierre
|
|
|
|
|
|
|
| |
|
|
|
|
| |
Hi Jen,
It might be important that everyone see what we want to show. Especially those who are doing the stain approach.
Do you want that we send the file to you ?
Pierre
|
|
|
|
|
|
|
| |
|
|
|
|
| |
Pierre
If you send a scaled down copy of any images you want to show on the site to me, then I can upload them to the server and embed them in a post.
|
|
|
|
|
|
|
 |
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum
|
|
|
|
|
|
|
|
|
| |
