-
I'm currently using a AMD Ryzen 5 7600 with a 4080super, if i want to SVP rife ai interpolate using 4.18v2 faster with more detail by transcoding rather than plying in real time, which cpu should I ge…
-
How to convert the model to ncnn? If there is no WARP layer in ncnn, then how can it be converted with a dimension of not 3 channels? in the inference function, the flow dimension has 2 channels inste…
-
When starting interpolation, it asks to enter the name of the model, which was not initially specified. When trying various options, it says "syntax error" or "model not defined". How and where to ent…
-
作者你好,我利用光流预测模型生成的光流图去约束rife生成的光流图,从结果上来看rife生成的光流图整体不错但是在细节方面不是很准确,我目前采用l1对光流计算损失,请问我可以采用其他什么损失能够提高生成光流图的准确性呢?
-
Will it be possible to deduplicate anime frames and interpolate it into full frames using the repository below?
[https://github.com/hyw-dev/AFI-ForwardDeduplicate](url)
-
Seems embarrassingly parallel, and there's an issue already to make it multithreaded. Why not make it CUDA-rific?
-
Currently the code will detect when the user is trying to use ensemble and a scale!=1.
It will then simply return 'currently not supported'.
Please add support for the combination of both paramete…
-
I downloaded all the downloads from https://github.com/AmusementClub/vs-mlrt/releases/tag/v14.test into "F:/Hybrid/64bit/vs-mlrt" and when calling:
```
# Imports
import vapoursynth as vs
import os…
-
Using `-s` has no effect (RIFE v4).
Timestep always defaults to 0.5.
-
This time, I've made optimization after a lot of trial and error.
The amendments are as follows
# 1. Add "scale_to_540, 675, 810"
Upscale_twice or 60 frame videos above 1080p almost unc…