ljchang commited on
Commit
5ed2577
·
verified ·
1 Parent(s): 1105a77

Clarify model licensing scope and dataset provenance

Browse files

Separate implementation licenses from pretrained-weight and training-data terms. Preserve existing valid grants and document unresolved distribution permissions. No weights or access settings changed.

Files changed (2) hide show
  1. LICENSE-NOTICE.md +9 -0
  2. README.md +11 -0
LICENSE-NOTICE.md ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ # svm_au — licensing scope notice
2
+
3
+ This distribution retains its existing `license: mit` metadata. Py-Feat's original software is MIT; that grant does not supply dataset-provider rights in the fitted classifiers, scalers or PCA parameters. A serialization-format update is not a new training provenance or permission grant.
4
+
5
+ The original AU-training provenance in the [2023 paper and supplementary materials](https://doi.org/10.1007/s42761-023-00191-4) names BP4D, BP4D+, DISFA, CK+ and UNBC-McMaster Shoulder Pain. Current documentation additionally identifies Aff-Wild2; the supplement's incomplete trailing dataset citation and separate EmotioNet discussion require a checkpoint-specific manifest rather than an assumed exhaustive list. DISFA+ is the paper's held-out AU evaluation source, not evidence of AU-detector training on that corpus.
6
+
7
+ These sources have separate access and use terms; see the [dataset register](https://github.com/cosanlab/py-feat/blob/main/LICENSES/DATASETS.md), [BP4D/BP4D+ provider page](https://www.cs.binghamton.edu/~lijun/Research/3DFE/3DFE_Analysis.html), and [Aff-Wild2 acquisition page](https://sites.google.com/view/dimitrioskollias/databases/aff-wild2). Parameter-set/model-sharing provisions and the agreement applicable at acquisition must be assessed for the particular checkpoint. Newer terms must not be assumed to apply retroactively.
8
+
9
+ This notice clarifies scope; it does not revoke existing valid grants or create a new license for third-party material. Dataset and teacher terms do not automatically relicense every trained artifact or inference output. A software license or model-card badge alone does not establish all checkpoint redistribution or commercial-use permissions.
README.md CHANGED
@@ -6,6 +6,8 @@ pipeline_tag: image-feature-extraction
6
 
7
  # svm_au
8
 
 
 
9
  ## Model Description
10
  svm_au combines histogram of oriented gradient feature extraction with a linear support vector machine to predict facial action units from single frame images.
11
 
@@ -129,4 +131,13 @@ print(pred)
129
 
130
  ```
131
 
 
 
 
 
 
 
 
 
 
132
 
 
6
 
7
  # svm_au
8
 
9
+ > **Licensing scope:** See the [license and provenance notice](LICENSE-NOTICE.md) before relying on this card's license metadata for pretrained-weight redistribution or commercial use. Existing valid grants are preserved.
10
+
11
  ## Model Description
12
  svm_au combines histogram of oriented gradient feature extraction with a linear support vector machine to predict facial action units from single frame images.
13
 
 
131
 
132
  ```
133
 
134
+ ## License and provenance
135
+
136
+ This distribution retains its existing `license: mit` metadata. Py-Feat's original software is MIT; that grant does not supply dataset-provider rights in the fitted classifiers, scalers or PCA parameters. A serialization-format update is not a new training provenance or permission grant.
137
+
138
+ The original AU-training provenance in the [2023 paper and supplementary materials](https://doi.org/10.1007/s42761-023-00191-4) names BP4D, BP4D+, DISFA, CK+ and UNBC-McMaster Shoulder Pain. Current documentation additionally identifies Aff-Wild2; the supplement's incomplete trailing dataset citation and separate EmotioNet discussion require a checkpoint-specific manifest rather than an assumed exhaustive list. DISFA+ is the paper's held-out AU evaluation source, not evidence of AU-detector training on that corpus.
139
+
140
+ These sources have separate access and use terms; see the [dataset register](https://github.com/cosanlab/py-feat/blob/main/LICENSES/DATASETS.md), [BP4D/BP4D+ provider page](https://www.cs.binghamton.edu/~lijun/Research/3DFE/3DFE_Analysis.html), and [Aff-Wild2 acquisition page](https://sites.google.com/view/dimitrioskollias/databases/aff-wild2). Parameter-set/model-sharing provisions and the agreement applicable at acquisition must be assessed for the particular checkpoint. Newer terms must not be assumed to apply retroactively.
141
+
142
+ This notice clarifies scope; it does not revoke existing valid grants or create a new license for third-party material. Dataset and teacher terms do not automatically relicense every trained artifact or inference output. A software license or model-card badge alone does not establish all checkpoint redistribution or commercial-use permissions.
143