除去tgmain.exe另一个选择——SR-BVPP的使用与对比

javplayer

使用方法

1.下载解压JAVPlayer+TecoGAN
2.下载
3.解压得到两个文件夹 BATCH 和 bvpp
以 JAVPlayer 1.12a 为例，上述文件夹放至 JAVP路径\TG 下



打开JAVP，随便拖入一个视频，先选择基础的模型（TG-STD或者TG-Z415）
然后就能选择到SR（SuperResolution）里的 SR-BVPP



对比

原视频



TG-STD+tgmain




TG-STD+SR-BVPP



TG-Z415+SR-BVPP



怎么说呢……我个人感觉因为有Z415的加持，后期的SR（SuperResolution）tgmain和bvpp效果差不多
具体还得看个人的喜好

Hellgate

bvpp 想要使用的过滤器无法在相应的计算机上工作。 Woohoo
(RTX 3xx 卡片)
使用 bvpp + REG6B 测试结果如何... 很好奇^^
或者设想bvpp + veai 1st prob 2std的话，会有很精彩的情况。

crosz

放进去选择不了

gdcc

学习了，谢谢

666666

这个方法我电脑一直报错  
JavPlayer_112a\TG\bvpp\torch\cuda\__init__.py:104: UserWarning:
NVIDIA GeForce RTX 3070 Laptop GPU with CUDA capability sm_86 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_37 sm_50 sm_60 sm_61 sm_70 sm_75 compute_37.
If you want to use the NVIDIA GeForce RTX 3070 Laptop GPU GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/

heyadai

按照说明放到对应目录下，无法调用
\JavPlayer_112\TG\bvpp>bvpp.exe basicvsr_plusplus_vimeo90k_bd.py basicvsr_plusplus_c64n7_8x1_300k_vimeo90k_bd_20210305-ab315ab1.pth ../input5 ../output5 --filename-tmpl {:04d}.png --device 0 --max-seq-len 300
GPU_ID=0
load checkpoint from local path: basicvsr_plusplus_c64n7_8x1_300k_vimeo90k_bd_20210305-ab315ab1.pth
Traceback (most recent call last):
  File "bvpp.py", line 77, in <module>
  File "bvpp.py", line 66, in main
  File "mmedit\apis\restoration_video_inference.py", line 127, in restoration_video_inference
  File "torch\nn\modules\module.py", line 727, in _call_impl
    result = self.forward(*input, **kwargs)
  File "mmcv\runner\fp16_utils.py", line 116, in new_func
  File "mmedit\models\restorers\basic_restorer.py", line 75, in forward
  File "mmedit\models\restorers\basicvsr.py", line 175, in forward_test
  File "torch\nn\modules\module.py", line 727, in _call_impl
    result = self.forward(*input, **kwargs)
  File "mmedit\models\backbones\sr_backbones\basicvsr_pp.py", line 348, in forward
  File "mmedit\models\backbones\sr_backbones\basicvsr_pp.py", line 218, in propagate
  File "torch\nn\modules\module.py", line 727, in _call_impl
    result = self.forward(*input, **kwargs)
  File "mmedit\models\backbones\sr_backbones\basicvsr_pp.py", line 434, in forward
  File "mmcv\ops\modulated_deform_conv.py", line 94, in forward
RuntimeError: CUDA error: no kernel image is available for execution on the device
[19672] Failed to execute script 'bvpp' due to unhandled exception!

E:\JavPlayer_112\TG\bvpp>pause
Press any key to continue . . .

Hellgate

https://twitter.com/Javski2/status/1556655860310429696

以下版本 Javplayer 1.12b 将创建并分发 RTX30XX 图形卡。
好消息啊。
请参考一下。

moonsun

heyadai 发表于 2022-8-8 20:11
按照说明放到对应目录下，无法调用
\JavPlayer_112\TG\bvpp>bvpp.exe basicvsr_plusplus_vimeo90k_bd.py ba ...

我也是同样的错误

imwangqing

看起来效果不错，试下我电脑能不能跑

weback

谢谢楼主分享