CoolFace
Datasetpublic

seonglae/vls-10k

VLS 10K 9,987 MS-COCO images paired with a long written description, a one-sentence spoken summary of that description, the spoken audio, and that audio pre-encoded by two neural codecs. Images and audio are embedded in the parquet, so the viewer renders them and one call opens the set: from datasets import load_dataset ds = load_dataset("seonglae/vls-10k", split="train") ds[0]["image"] # PIL image ds[0]["audio"] # decoded waveform ds[0]["sst"] # the sentence that was… See the full description on the dataset page: https://huggingface.co/datasets/seonglae/vls-10k.

sourceHugging Faceupdated 1mo agoView on Hugging Face
0likes82downloads
../
filetest-00000.parquet404.8 MBdownload
filetest-00001.parquet173.0 MBdownload
filetrain-00000.parquet404.6 MBdownload
filetrain-00001.parquet406.0 MBdownload
filetrain-00002.parquet405.4 MBdownload
filetrain-00003.parquet407.3 MBdownload
filetrain-00004.parquet405.5 MBdownload
filetrain-00005.parquet405.1 MBdownload
filetrain-00006.parquet406.0 MBdownload
filetrain-00007.parquet406.3 MBdownload
filetrain-00008.parquet406.3 MBdownload
filetrain-00009.parquet403.8 MBdownload
filetrain-00010.parquet403.4 MBdownload
filetrain-00011.parquet166.8 MBdownload
filevalidation-00000.parquet403.8 MBdownload
filevalidation-00001.parquet172.1 MBdownload

seonglae/vls-10k · main · files are served by the source, never re-hosted here