Mirrors/ECCV2022-RIFE
1
0
Fork 0
mirror of https://github.com/hzwer/ECCV2022-RIFE.git synced 2025-12-16 16:37:51 +01:00
Code Issues Packages Projects Releases Wiki Activity

Update README.md

Browse Source
This commit is contained in:
hzwer
2020-12-02 19:12:48 +08:00
committed by GitHub
parent 8165ad30ef
commit e6c5e8b872
1 changed files with 5 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
README.md
View File

@@ -71,9 +71,7 @@ ffmpeg -r 10 -f image2 -i output/img%d.png -s 448x256 -vf "split[s0][s1];[s0]pal
```
## Evaluation
First, you should download [RIFE model reported by our paper](https://drive.google.com/file/d/1c1R7iF-ypN6USo-D2YH_ORtaH3tukSlo/view?usp=sharing).
We will release our training and benchmark validation code soon.
Download [RIFE model reported by our paper](https://drive.google.com/file/d/1c1R7iF-ypN6USo-D2YH_ORtaH3tukSlo/view?usp=sharing).
**Vimeo90K**: Download [Vimeo90K dataset](http://toflow.csail.mit.edu/) at ./vimeo_interp_test
@@ -85,12 +83,14 @@ python3 benchmark/MiddelBury_Other.py
# (Final result: "2.058")
```
## Training
## Training and Reproduction
Because Vimeo90K dataset and the corresponding optical flow labels are too large, we cannot provide a complete dataset download link. We provide you with [a subset containing 100 samples](https://drive.google.com/file/d/1_MQmFWqaptBuEbsV2tmbqFsxmxMIqYDU/view?usp=sharing) for testing the pipeline. Please unzip it at ./dataset
Each sample includes images (I0 I1 Imid : 9 x 256 x 448), and optical flow (flow_t0, flow_t1: 4, 256, 448).
For origin images, you can download them from [Vimeo90K dataset](http://toflow.csail.mit.edu/).
For generating optical flow labels, our paper use [pytorch-liteflownet](https://github.com/sniklaus/pytorch-liteflownet). We also recommend [RAFT](https://github.com/princeton-vl/RAFT) because it's easier to configure. We recommend generating optical flow labels on 2X size images for higher quality. You can also generate labels during training, or finetune the optical flow network on the training set. The final impact of the above operations on Vimeo90K PSNR is expected to be within 0.3.
For generating optical flow labels, our paper use [pytorch-liteflownet](https://github.com/sniklaus/pytorch-liteflownet). We also recommend [RAFT](https://github.com/princeton-vl/RAFT) because it's easier to configure. We recommend generating optical flow labels on 2X size images for better labels. You can also generate labels during training, or finetune the optical flow network on the training set. The final impact of the above operations on Vimeo90K PSNR is expected to be within 0.3.
We use 16 CPUs, 4 GPUs and 20G memory for training:
```