Add Q-Align splits for KADID10k and SPAQ

README.md

@@ -6,6 +6,21 @@ license: cc-by-sa-4.0
 
 This repo contains the meta information of datasets stored in [chaofengc/IQA-PyTorch-Weights](https://huggingface.co/chaofengc/IQA-PyTorch-Weights). They are used in the training codes of the [pyiqa toolbox](https://github.com/chaofengc/IQA-PyTorch).
 
+# Q-Align split annotations
+
+The `meta_info_KADID10kDataset.csv` and `meta_info_SPAQDataset.csv` files
+include a `qalign_split` column. Rows present in the corresponding Q-Align test
+manifest are marked `test`; all remaining rows are marked `train` as the
+complement of that test manifest. Image names were matched against the
+Q-Align manifests. These columns provide a consistent Q-Align test split
+alongside the existing HF train/validation folds.
+
+The Q-Align test manifests are available for these datasets, but a separate
+Q-Align training manifest was not available when these annotations were
+prepared. Therefore, `train` means “not present in the supplied Q-Align test
+manifest” and should not be interpreted as confirmation that every such row
+was used to train Q-Align.
+
 # Disclaimer for Datasets Included
 
 This collection of datasets is compiled and maintained for academic, research, and educational purposes. It is important to note the following points regarding the datasets included in this Collection: