HMQC zones picking (manual) leads to crash

[nes-cgn]

In the following example (HMQC from nmr-dataset1 case 2591-86-8), automatic zones picking does not allow a meaningful result. So, the picking needs to be done manually. And in that case, the three signals in the upper right corner of the correlation (see expansion 2 from attached files) always leads to a crash, if I try to select the three individual cross peaks for zone definition...
hmqc-light.zip

[hamed-musallam]

@nes-cgn
can you try the last version....

[nes-cgn]

I tried the latest version, problem continues. The notebook is quite powerful, I do not think it is a processor issue.

[hamed-musallam]

@lpatiny @jobo322
this bug related to nmr-processing package

[jobo322]

@nes-cgn could you check if this was improved?

[nes-cgn]

I tried, and what I see is that it does not immediately crash. This is an advantage. However, when I try to do an automated picking, I get what you see in the file called "...-new":

If I do remove the automatically picked peaks and define the zones manually, the frames for the zones are visible, as visible in the file "...-new2":

If I define the entire group of the three "difficult" cross peaks, it somehow finds the maxima, but I would prefer that I can select the region and then, only one (centered) peak is defined. As in the "all-in-one" approach, the three different signals fall into the same zone which I find less useful:

[jobo322]

I am looking for a solution. I found that we pass to the method all the intensity range. we could handle the problem with noiseFactor e.g for auto detection with noiseFactor equal to 5 I got the next picture

it is not still the best result but it is better, I am trying to understand why the method depends in this way of noiseFactor

[nes-cgn]

Well, as I mentioned - it is primitive, but if you look at the "-new2" picture three graphs above, in the worst of all cases, one could define an option that sets the centre of a (manually defined) region as the picked peak. That is by no means automatic recognition, but at least, like that you could add a value in critical cases where other attempts fail

[jobo322]

Apparently, this problem is fixed, could you have a look?