CoolFace
Datasetpublic

jakubkrapiec/crf-vmaf-training-data

CRF <-> VMAF training data Training datasets behind Alice-1 - a LightGBM model that predicts the CRF value needed to hit a target VMAF score for a given video segment, codec and resolution. Contents File Rows Description training_table.parquet 1,762,232 Main table. Feature aggregates -> CRF for a target VMAF, per codec/resolution. probe_data.parquet 52,316 Probe-encode data. Two 2s probe encodes per (segment, codec, resolution) cell with measured VMAF… See the full description on the dataset page: https://huggingface.co/datasets/jakubkrapiec/crf-vmaf-training-data.

sourceHugging Facecc0-1.0updated 24d agoView on Hugging Face
0likes45downloads
Dataset Card

CRF <-> VMAF training data

Training datasets behind Alice-1 - a LightGBM model that predicts the CRF value needed to hit a target VMAF score for a given video segment, codec and resolution.

Contents

FileRowsDescription
training_table.parquet1,762,232Main table. Feature aggregates -> CRF for a target VMAF, per codec/resolution.
probe_data.parquet52,316Probe-encode data. Two 2s probe encodes per (segment, codec, resolution) cell with measured VMAF + bitrate.

Schema: training_table

One row = (video segment codec target resolution * target VMAF) -> the CRF that achieves that VMAF.

ColumnTypeDescription
source_keystringSource video identifier (dataset/filename)
datasetstringCorpus subset (main, hc, ext, …)
splitstringTrain/validation/test split
segment_startfloatSegment start time (s)
codecstringx264, x265, vp9, av1
presetstringEncoder preset (mostly veryfast; null for some rows)
source_width / source_heightintSource resolution
target_width / target_heightintEncoded resolution (720/1080/1440/2160)
fpsfloatFrames per second
segment_durationfloatSegment length (s, ~10)
vmaf_modelstringVMAF model version
fit_methodstringHow CRF was extracted from the encode ladder (logistic fit)
si_mean / si_stdfloatSpatial Information (ITU-T P.910) mean/std over segment
ti_mean / ti_stdfloatTemporal Information mean/std
vmafmotionfloatVMAF motion feature
target_vmafintTarget VMAF score (60–95)
crffloatLabel: CRF achieving target_vmaf

Schema: probe_data

One row = (segment codec resolution) with two cheap probe encodes.

ColumnDescription
probe_crf1, probe_vmaf1, probe_bitrate1_kbpsFirst probe encode (2s): CRF, measured VMAF, bitrate
probe_crf2, probe_vmaf2, probe_bitrate2_kbpsSecond probe encode (different CRF)
probe_secProbe duration (2.0s)
probe_vmaf= probe_vmaf1 (primary probe feature)
probe_slopeVMAF slope between the two probe points
probe_log_br, probe_log_br2log10 of probe bitrates

Sources

~40,861 unique (segment, codec, resolution) cells from ~9,000 source clips: Blender Open Movies, Internet Archive public-domain films, NASA SVS, Wikimedia Commons, Pexels/Pixabay/Coverr/Mixkit stock, Xiph.org test media, YouTube CC. All sources are openly licensed.

Citation

If you use this dataset, please link the repo.

License

CC0-1.0 (public domain dedication). Feature aggregates and measurements computed by the authors from openly licensed sources; no video content is included.