Closed smaranjitghose closed 4 years ago
I would like to work on this issue. Please assign it to me
@annu12340 has done the preliminary work to build on 👌
For the next phase of the issue, I want you to modify the script or add another one so that if I have an image say img1
and there's another image say img2
which is actually a part of the first image img1
(consider I have a fingerprint as my first image and a part(say quadrant) of the fingerprint as the second image), our script should be able to identify this with a score and also draw a bounding box over the original image img1
that tells us which part was taken to form img2
Hi @smaranjitghose,
I was trying to understand, the next phase of this issue. I tried writing a script for it.
I wanted to confirm, if I am understanding the problem statement correctly,
I am getting this output..is this how it should be?
Hey Purva, I had modified the model, to detect the difference between the 2 images and to highlight these difference also
Ohh okay😂😂 I thought the next phase was, To find whether a small image was a subset of another large image. And then locate the smaller image in the larger image, and draw a bounding box. I think I messed up understanding the problem statement😅 Sorry!
Purva, I am not sure about that. Just thought of informing you about the update :smile:
@purva98 I think this might work for a lot of cases, can you please share the link of the Colab Notebook
@annu12340 could you please share a snapshot of your work or the main highlight of you script
Hello @smaranjitghose , the link of the small script is https://colab.research.google.com/drive/1ptEj1hWNAec_Nfn8Zp9B2e1cgrwlYx2Y Please let me know, if I should also include : Detection of multiple instances of the subset in larger image.
Yes, this is really nice And you can start working for detecting multiple instances too
hello @smaranjitghose , in this file, i have implementation of single instance detection, followed by multiple instance detection, https://colab.research.google.com/drive/1ptEj1hWNAec_Nfn8Zp9B2e1cgrwlYx2Y should I make a script and do a PR?
Yes, this works well!
Go ahead and give a PR for this and continue with the script.
Also, make sure you use a separate directory for this say img_subset
Task
Compare two input images and return a value that tells how visually similar the given images are. The lower the score, the more contextually similar the two images are with a score of '0' being identical.