NAKASASA / getnative

Find the native resolution(s) of upscaled material (mostly anime)

Getnative

Find the native resolution(s) of upscaled material (mostly anime)

Usage

Start with installing the depdencies through pip.

Start by executing:

$ python getnative.py inputFile [--args]

That's it.

Requirements

To run this script you will need:

• Python 3.6
• matplotlib
• Vapoursynth R39+
• descale (really slow for descale) or descale_getnative (perfect for getnative)
• ffms2 or lsmash or imwri

Example Output

Input Command:

$ python getnative.py /home/infi/mpv-shot0001.png -k bicubic -b 1/3 -c 1/3

Output Text:

Using imwri as source filter
501/501
Kernel: bicubic AR: 1.78 B: 0.33 C: 0.33
Native resolution(s) (best guess): 720p, 987p
done in 29.07s

Output Graph:

Args

Property	Description	Default value	Type
help	Automatically render the usage information when running -h or --help	False	Action
	Absolute or relative path to the input file	Required	String
frame	Specify a frame for the analysis.	num_frames//3	Int
mode	Choose a predefined mode ["bilinear", "bicubic", "bl-bc", "all"]	None	String
kernel	Resize kernel to be used	bicubic	String
bicubic-b	B parameter of bicubic resize	1/3	Float
bicubic-c	C parameter of bicubic resize	1/3	Float
lanczos-taps	Taps parameter of lanczos resize	3	Int
aspect-ratio	Force aspect ratio. Only useful for anamorphic input	w/h	Float
min-height	Minimum height to consider	500	Int
max-height	Maximum height to consider	1000	Int
use	Use specified source filter (e.g. "lsmas.LWLibavSource")	None	String
is-image	Force image input	False	Action
generate-images	Save detail mask as png	False	Action
plot-scaling	Scaling of the y axis. Can be "linear" or "log"	log	String
plot-format	Format of the output image. Specify multiple formats separated by commas. Can be svg, png, tif(f), and more	svg	String
show-plot-gui	Show an interactive plot gui window.	False	Action
no-save	Do not save files to disk.	False	Action

Warning

This script's success rate is far from perfect. If possible, do multiple tests on different frames from the same source. Bright scenes generally yield the most accurate results. Graphs tend to have multiple notches, so the script's assumed resolution may be incorrect. Also, due to the current implementation of the autoguess, it is not possible for the script to automatically recognize 1080p productions. Use your eyes or anibin if necessary.

Thanks

BluBb_mADe, kageru, FichteFoll, stux!

Help?

Join https://discord.gg/V5vaWwr (Ask in #encode-autism for help)