Skip to content

Instantly share code, notes, and snippets.

View WolframRhodium's full-sized avatar
🎯
Focusing

WolframRhodium

🎯
Focusing
View GitHub Profile
import vapoursynth as vs
from vapoursynth import core as _vscore
class _Plugin:
def __init__(self, namespace):
self.__dict__.update((name, getattr(namespace, name)) for name in dir(namespace)) # func_name : func
class _Core:
def __init__(self):
self.__dict__.update((name, get_plugin(name)) for name in dir(_vscore)) # (namespace : (func_name : func)) or (attr_name : attr)
@WolframRhodium
WolframRhodium / Usage.md
Last active August 27, 2023 16:31
WaDIQaM calculator for VapourSynth

Official implementation based on Chainer (requires CUDA)

from vapoursynth import core
import chainer
# chainer.global_config.cudnn_deterministic = False

from vs_wadiqam_chainer import wadiqam_fr, wadiqam_nr


model_folder_path = "deepIQA-master\models" # path to the folder that contains model's parameter files
@WolframRhodium
WolframRhodium / impulse_cheatsheet.md
Last active December 30, 2018 13:55
fmtc.resample(kernel='impulse') cheatsheet
fmtconv equivalent remarks
impulse=[1, 1, 1], fv=-1, fh=-1 std.Convolution([1]*9) or rgvs.RemoveGrain(20) Radius=1 box filtering. fv and fh are required to force the processing.
scale=1/2, kernel='box' scale=1/2, kernel='impulse', impulse=[1]*3, kovrspl=3 Box downscaling with a factor of 2
scale=1/4, kernel='box' scale=1/4, kernel='impulse', impulse=[1]*5, kovrspl=5 Box downscaling with a factor of 4
scale=1/2, kernel='bilinear' scale=1/2, impulse=[0.5, 1, 0.5], kovrspl=2 Bilinear downscaling with a factor of 2
scale=1/3, kernel='bilinear' scale=1/3, impulse=[0.5, 1, 0.5], kovrspl=2 Bilinear downscaling with a factor of 3
scale=1/4, kernel='bilinear' scale=1/4, impulse=[0.5, 1, 0.5], kovrspl=2 Bilinear downscaling with a factor of 4
@WolframRhodium
WolframRhodium / resizer_speed.md
Last active December 16, 2018 05:20
fmtconv vs zimg (resize)

Environment:

Intel Core i7-6700HQ @ 2.60GHz, 8 GB DDR4-2133 x2,

Windows 10 1809, VapourSynth R44 64bit, fmtconv r20(CRC: 6EEDD665)

Result:

Input dimensions Output dimensions Format Backend FPS
@WolframRhodium
WolframRhodium / AOD_net-symbol.json
Last active October 31, 2018 09:47
AOD_net (fix preprocessing)
{
"nodes": [
{
"op": "null",
"name": "data",
"inputs": []
},
{
"op": "_plus_scalar",
"name": "_plusscalar0",
@WolframRhodium
WolframRhodium / -Publications.md
Last active January 17, 2018 19:42
Publications

Muonium

章节

  1. KNLMeansCL 参数介绍
  2. 从Sobel算子到ringmask2()
  3. iterLineDarken —— 一种新型迭代式线条加深方法
  4. 两种新式图像结构迁移方法
  5. MinBlurMod —— 一种新式去点状晕轮方法
@WolframRhodium
WolframRhodium / numerical_analysis.py
Last active June 18, 2017 15:38
Implementation of numerical analysis algorithms in Python
import numpy as np
import scipy as sp
from scipy import misc
import sympy as sym
import math
# Helper functions
# https://en.wikipedia.org/wiki/Norm_(mathematics)
def calculate_error(diff, norm=0):
# Parameters
input = YUV420P16
# matrix = "709"
matrix = mvf.GetMatrix(input, matrix, True)
process_y = True
sigma_y = 2.0
basic_y_args = dict(profile="fast", group_size=8, bm_range=6)
process_uv = False