-
Models:
69 = rife-v4.25 (ensemble=False)
70 = rife-v4.25-lite (ensemble=False)
71 = rife-v4.25-heavy (ensemble=False)
72 = rife-v4.26 (ensemble=False)
all cause artifacts:
![grafik](https://gith…
Selur updated
3 weeks ago
-
I downloaded the rife v2 version of the onnx model you provided, do you have the code on how to convert it to an onnx model? And would also like to ask you about how to use the converted onnx, do you …
-
@hzwer you say that:
**Currently, it is recommended to choose 4.26 by default for most scenes.**
FYI. The impression I'm getting from SVP forum and from personal experience is that v4.25 might sti…
-
作者您好,在RIFE的inference_img.py中,是这样处理输入帧:
https://github.com/hzwer/ECCV2022-RIFE/blob/638322e8bfe6daca4439f7b56503fef4b090d34f/inference_img.py#L57-L60
在Practical-RIFE中,输入帧被resize到(448, 256):
https:…
ZXMMD updated
2 weeks ago
-
作者您好,请问您尝试过使用高分辨率数据集(比如[X-TRAIN](https://github.com/JihyongOh/XVFI))训练RIFE吗?我在训练过程中遇到了一些问题。
实验设置如下:
1. 构造三帧组。X-TRAIN数据集中一个视频有65帧(索引为0到64),可以构造不同时间间隔的三帧组:0,1,2; 0,2,4; 0,3,6; 0,4,8; 0,5,10; ...; 0,32…
-
I won't attach my test image because it's too weird, but I assume any 1024x1024 image will do the trick. In my case I used a .png image.
resized_img = img.resize((new_width, new_height), Image.AN…
-
你好作者,这是一个非常棒的工作,我想和您请教沟通下如何对模型做轻量化的部署。
1. 降低ResConv层数及channel。
2. 对上下采样做更快的,例如使用conv替换。
3. 估计1/2,1/4甚至1/8的光流,上采样至原分辨率做warp
4. grid_sample操作较为耗时,是否有更好的替换方案?可不可以不内建warp
5. 为了解决4,是否可以在不做warp的情况下做多…
-
This part causing error
```
def pad_image(img, scale):
_, _, h, w = img.shape
tmp = max(32, int(32 / scale))
ph = ((h - 1) // tmp + 1) * tmp
pw = ((w - 1) // tmp + 1) * tmp
…
-
Is it possible to solve this problem yourself and are there plans to further update flowframes?
![352762150-ee9a19e1-aa95-481f-861c-567f54e4ded1](https://github.com/user-attachments/assets/b1300596-e…
-
```
Failed to evaluate the script:
Python exception: vsmlrt.RIFEMerge: tile size must be divisible by 64 (192, 160)
Traceback (most recent call last):
File "src\\cython\\vapoursynth.pyx", line 3…