|
|
--- |
|
|
license: mit |
|
|
task_categories: |
|
|
- image-feature-extraction |
|
|
- video-classification |
|
|
language: |
|
|
- en |
|
|
--- |
|
|
|
|
|
# Generating_Video_Data |
|
|
|
|
|
``` |
|
|
Hanzhe Liang |
|
|
modified: 23 May |
|
|
``` |
|
|
|
|
|
😊 This library contains some common data for training video generation models. This is a library of summary datasets, with thanks to the original authors for their contributions. Regardless of which data you need, the following steps: |
|
|
```bash |
|
|
mkdir data |
|
|
cd data |
|
|
``` |
|
|
Next, go to my cloud database link and download the data you want into the folder```data/```. |
|
|
Once you've downloaded the data you need, we've found preprocessing steps from its official implementation, which you can follow to process the data. |
|
|
For memory requirements, we provide the files as single zip files, you need to extract the file named data_name.rar with the unzip command: ```unrar x data_name.rar``` |
|
|
|
|
|
---- |
|
|
|
|
|
### Something Something V2 |
|
|
Please check the data path. |
|
|
```bash |
|
|
cd somethingv2 |
|
|
python extract_frames.py |
|
|
python process_somethingv2.py |
|
|
``` |
|
|
You can check data list ```somethingv2/*.txt``` |
|
|
|
|
|
### Human3.6M |
|
|
I provided two datasets, both processed and unprocessed. |
|
|
|
|
|
Unprocessed |
|
|
```bash |
|
|
cd human36m/human36m_unprocessed |
|
|
python build_clip_dataset.py |
|
|
python make_list.py |
|
|
``` |
|
|
You can check data list ```human36m/human36m_unprocessed/*.txt``` |
|
|
|
|
|
or you can directly check the processed data in ```human36m/human36m_processed``` |
|
|
|
|
|
### YouTubeDriving |
|
|
```bash |
|
|
cd youtubedriving |
|
|
python make_list.py |
|
|
``` |
|
|
You can check data list ```youtubedriving/*.txt``` |
