|
|
--- |
|
|
license: cc-by-4.0 |
|
|
--- |
|
|
|
|
|
For IQA and IAA, we use AVA, TAD66K, Koniq-10k, livec and spaq datasets, which are captured in the **real world**. For UGC assessment, we propose the RealQA dataset. |
|
|
📚 All image related data can be loaded in this repo. |
|
|
📚 The training and test conversation template can be found in the [hugging face](https://huggingface.co/datasets/MingxingLi/RealQA-conversation). |
|
|
|
|
|
|
|
|
```bash |
|
|
unzip dataset_split.zip |
|
|
``` |
|
|
|
|
|
> The unzip program will intelligently recognize that dataset_split.zip is the first part of a multi-volume compressed package, and then it will automatically search for subsequent volumes such as dataset_split.z01 in the same directory, merge them together, and finally decompress the original complete dataset.zip file. |
