Donfa1con/talking-head-metrics

Talking-Head Metrics SI-184 Evaluation Dataset

What this is

This dataset holds the material to score talking-head listener generation on the SI-184 benchmark. SI-184 is a set of 184 dyadic conversation clips. Seven talking-head systems each generate a listener video for every clip. This dataset gives you those generated videos. It also gives you the extracted features that the metrics read.

Six feature extractors run on every video:

• EMOCA — FLAME expression and pose codes.
• LivePortrait — implicit keypoint motion.
• InceptionV3 — image features for FID.
• I3D — video features for FVD.
• ArcFace / insightface — face-identity embeddings.
• SyncNet — audio-visual sync embeddings.

GT also carries VAD (voice-activity) features. VAD feeds the listening variant of the motion metrics (rPCC, TLCC, P-FD, SID, Variance), which scores the listener only over the segments where the partner speaks.

With these files you compute nine metrics without generating the videos again and without running the extractors again:

• rPCC — reference Pearson correlation coefficient.
• TLCC — time-lagged cross-correlation.
• P-FD — pose Fréchet distance.
• SID — sample inception distance.
• Variance — motion variance.
• FID — Fréchet inception distance.
• FVD — Fréchet video distance.
• CSIM — face-identity cosine similarity.
• LSE-C / LSE-D — lip-sync error (confidence and distance).

The 184 clip identifiers are listed in pairs184.txt. Column 1 gives the listener stem. Column 2 gives the partner (speaker) stem.

Systems

The dataset covers eight sources. GT is the ground truth. It carries features only, not video. The other seven are the generated systems. Each generated system carries video plus features.

source	role	video	features
GT	ground truth	no	emoca, liveportrait, inception, i3d, insightface, syncnet_emb, vad
xxxx-1	generated (dyadic)	yes	emoca, liveportrait, inception, i3d, insightface, syncnet_emb
AvatarForcing	generated (dyadic)	yes	emoca, liveportrait, inception, i3d, insightface, syncnet_emb
dystream	generated (dyadic)	yes	emoca, liveportrait, inception, i3d, insightface, syncnet_emb
FLOAT	generated (lip-sync)	yes	emoca, liveportrait, inception, i3d, insightface, syncnet_emb
ditto	generated (lip-sync)	yes	emoca, liveportrait, inception, i3d, insightface, syncnet_emb
SoulX_FlashHead_Lite	generated (lip-sync)	yes	emoca, liveportrait, inception, i3d, insightface, syncnet_emb
SoulX_FlashHead_Pro	generated (lip-sync)	yes	emoca, liveportrait, inception, i3d, insightface, syncnet_emb

Layout

pairs184.txt                                   # 184 lines: <listener>.wav <speaker>.wav
README.md                                      # this card
GT/                                            # features only, no video
  LS_AL_emoca/<stem>.npz                        # 184  EMOCA codes
  LS_AL_liveportrait/<stem>.npz                 # 184  LivePortrait motion
  LS_AL_inception/<stem>.npy                     # 184  InceptionV3 features (FID)
  LS_AL_i3d/<stem>.npy                           # 184  I3D features (FVD)
  LS_AL_insightface/<stem>.npz                   # 184  ArcFace embeddings (CSIM)
  LS_AL_syncnet_emb/<stem>.npz                   # 184  SyncNet embeddings (LSE)
  LS_AL_vad/<stem>.npz                           # 184  voice-activity (listening variant of motion metrics)
<model>/                                        # each of the 7 generated systems
  LS_AL/<stem>.mp4                              # 184  generated listener video (25 fps)
  LS_AL_emoca/<stem>.npz                        # 184  EMOCA codes
  LS_AL_liveportrait/<stem>.npz                 # 184  LivePortrait motion
  LS_AL_inception/<stem>.npy                     # 184  InceptionV3 features
  LS_AL_i3d/<stem>.npy                           # 184  I3D features
  LS_AL_insightface/<stem>.npz                   # 184  ArcFace embeddings
  LS_AL_syncnet_emb/<stem>.npz                   # 184  SyncNet embeddings

Every directory holds exactly 184 files. Every stem matches column 1 of pairs184.txt. Feature file types: emoca, liveportrait, insightface and syncnet_emb are .npz; inception and i3d are .npy.

Licenses

Use this dataset for non-commercial purposes only.

The dataset as a whole is licensed CC BY-NC 4.0.

The GT/* features are derived from the Seamless Interaction dataset (Meta / FAIR). They are attributed to Seamless Interaction and are covered by CC BY-NC 4.0.

Each generated system's videos stay under that system's own upstream license. The list below states each license as declared by its upstream project:

source	upstream license of the generated videos
ditto	Apache License 2.0
FLOAT	CC BY-NC 4.0 (Attribution-NonCommercial 4.0 International)
SoulX_FlashHead_Lite	Apache License 2.0
SoulX_FlashHead_Pro	Apache License 2.0
AvatarForcing	CC BY-NC 4.0 (Attribution-NonCommercial 4.0 International)
dystream	No license file is declared by the upstream project. Treat the videos as research use only. All rights stay with the original authors.
xxxx-1	Non-commercial research only. The model weights are under a non-commercial research licence (LICENSE-MODEL.md). The renderer and streamer code are PolyForm Noncommercial (LICENSE-RENDERER.md / LICENSE-STREAMER.md). Do not redistribute the weights.

The strictest term applies to any part you reuse. Several systems are non-commercial or research-only, so the combined dataset is non-commercial.

Not included

The dataset does not include the ground-truth raw video or the ground-truth audio. Those come from the Seamless Interaction release, and you must accept its access terms to obtain them.

Fetch the ground-truth video and audio with the code repository script data/download_si184.py. That script writes:

GT/LS_AL/<stem>.mp4        # ground-truth video (both dyad roles)
GT/LS_AL_wav/<stem>.wav    # driving / partner audio (both dyad roles)

Usage

1. Get the talking-head-metrics code repository.

2. Copy this whole tree into the repository at data/. Merge it with the existing data/ content (pairs184.txt and the download_si184.py script).

3. Run data/download_si184.py to add the ground-truth video and audio (see "Not included").

4. Compute every metric:

   bash pipeline/4_metrics.sh
   

That script reads the features and videos in this tree. It does not re-generate the videos. It does not re-extract the features.