Invelos Forums->DVD Profiler: Plugins |
Page:
1 2  Previous Next
|
|
Tool: SharpnessCheck |
|
|
|
|
|
| Author |
Message |
| Registered: October 22, 2015 | Reputation:  |  Posts: 311 |
| | Posted: | | | | SharpnessCheck ver. 1.1.0 (64-bit)
Scan 3920 profiles on Intel (Coffee Lake) i5-8600K CPU @ 3.60GHz (6 cores, no hyperthreading):
Laplacian:
(Front)
parallel tasks
6 ==> 21.86 secs
3 ==> 30.89 secs
2 ==> 43.94 secs
1 ==> 84.25 secs
(Back)
6 ==> 22.23 secs
3 ==> 30.97 secs
2 ==> 44.28 secs
1 ==> 84.46 secs
Wavelet:
(Front)
parallel tasks
6 ==> 13.18 secs
3 ==> 14.78 secs
2 ==> 17.58 secs
1 ==> 29.13 secs
(Back)
6 ==> 13.31 secs
3 ==> 14.93 secs
2 ==> 18.01 secs
1 ==> 30.09 secs
Windows resource monitor shows all processors are active, no matter how many parallel tasks are configured.
All cores at 100% when parallel tasks at 6 or more, and all cores at a lower % of activity depending on the reduced number of parallel tasks configured.
Hope that helps. |
| | Registered: March 14, 2007 | | Reputation: |  Posts: 4,840 |
| | Posted: | | | | Thanks ObiKen!
Interesting results! I hadn't actually timed Laplacian vs. Wavelet. I saw that Wavelet was faster, but I didn't realize that it was that much faster. Anyway, parallel processing really speeds up things when applied to suitable tasks.
For those who wish to find cover scans in their collection that are candidates for re-scanning, my tool CoverInfo is - in my opinion - a better place to start than SharpnessCheck. The physical attributes tell a great deal about the cover scans. It's really after one has addressed the obvious candidates there that it's worth trying to find blurry scans with SharpnessCheck. I would wish that it was more effective than it is in that regard. Still, it is not totally without merit.
Have you compared the results to see which function that seems best in finding blurry scans? Even though the results are different, I find it hard to say that one is more efficient than the other. How about you?
One program that I think could really benefit from parallel processing is Database Repair. I wish I could write my own version, but unfortunately I neither have the info nor the tools to do so. | | | My freeware tools for DVD Profiler users.
Gunnar |
| | Registered: October 22, 2015 | | Reputation: | | Posts: 311 |
| | Posted: | | | | Quoting GSyren: Quote:
Have you compared the results to see which function that seems best in finding blurry scans? Even though the results are different, I find it hard to say that one is more efficient than the other. How about you?
There was a reasonable correlation between the two methods. Those with poor contrast and /or fewer colours were typically rated with lower variance. Back covers tended to have higher variance values than their corresponding front cover, as they had, in the main, more information (text, colours, images, contrast). Either method is fine. The output result can be sorted from lowest to highest variance (that is, worst to best), so allows for quick identification and confirmation that the image in question is in need of fixing. In some cases, a low variance image cannot be improved because it is mainly a dark image (or solid colour) with small variations in pixel intensities. |
| | Registered: January 26, 2009 |  Posts: 186 |
| | Posted: | | | | Good explanation, but I think my issue maybe that newer Intel CPU's have E (efficiently) and P (performance) cores. Do not know if you are running one, but may be a factor in all this on newer CPU's.
And my i9 14900KF has 16 E and 8 P cores. Also does not have graphics on cpu. | | | | Last edited: by specise_8472 |
| | Registered: March 14, 2007 | | Reputation: | | Posts: 4,840 |
| | Posted: | | | | Quoting specise_8472: Quote:
Good explanation, but I think my issue maybe that newer Intel CPU's have E (efficiently) and P (performance) cores. Do not know if you are running one, but may be a factor in all this on newer CPU's.
And my i9 14900KF has 16 E and 8 P cores. Also does not have graphics on cpu. Right, and if I understand correctly, the P cores can run two threads and the E cores can run a single thread, so that would be 16 + 16, making it a total maximum of 32 threads. That’s a pretty powerful PC that you have.  | | | My freeware tools for DVD Profiler users.
Gunnar |
|
|
|
Invelos Forums->DVD Profiler: Plugins |
Page:
1 2 Previous Next
|
|
|