Differences

This shows you the differences between two versions of the page.

--- howto:stashface [2025/04/19 21:42] – created Wulf Rajek
+++ howto:stashface [2025/04/20 13:08] (current) – Wulf Rajek
@@ Line 1: / Line 1: @@
 ====== Stashface ======
+Stashface is a Gradio web user interface with api endpoints for facial recognition, primarily for use with the Stash plugin Visage. It requires a vector database of the faces with references to the stashdb uuids and file with uuid to metadata mapping. A password protected zip file of the mapping is in the docker image and huggingface git source. An own database can be created using visage-ml.
+The performers.json mapping file should be in the following format for stashface:
+<code - performers.json>
+{
+"xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx": {"name": "Firstname Lastname", "image": "https://cdn.example.org/images/uuid", "country": "GB"},
+"xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx": {"name": "Firstname2 Lastname2", "image": "https://cdn.example.org/images/uuid", "country": "HU"},
+"xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx": {"name": "Firstname3 Lastname3", "image": "https://cdn.example.org/images/uuid", "country": "US"},
+...
+}
+</code>
 https://huggingface.co/spaces/cc1234/stashface
@@ Line 66: / Line 79: @@
-Deprecated Visage version:
+Deprecated Visage userscript version:
 https://github.com/cc1234475/visage
@@ Line 77: / Line 90: @@
       - "https://cc1234-stashface.hf.space"</code>
-Visage-ML:
+====== Visage-ML ======
+Visage-ML is used to create a database from images for use with stashface/visage plugin.
+The directory structure of the images needs to be in the following format:
+<code>
+person_name
+  - image_file_001.jpeg
+  - another_01.jpeg
+</code>
+Updated fork:
 https://github.com/badde57/visage-ml
+Original:
+https://github.com/cc1234475/visage-ml
+<code>
+git clone https://github.com/badde57/visage-ml
+#Hashing:
+cd visage-ml/hasher/
+mkdir weights
+cd weights/
+wget https://github.com/serengil/deepface_models/releases/download/v1.0/facenet512_weights.h5
+wget https://github.com/serengil/deepface_models/releases/download/v1.0/retinaface.h5
+cd ..
+mkdir -p /opt/visage/images
+vi docker-compose.yaml
+# adjust image folder in compose
+docker-compose build
+docker-compose up
+#builder
+cd ../builder
+vi docker-compose.yaml
+# adjust image folder in compose
+docker-compose build
+docker-compose up
+#matcher
+cd ../matcher
+cp /opt/visage/images/face.db .
+cp /opt/visage/images/face.json .
+vi performers.json
+{
+"93aa1a98-e86e-47a2-b661-f5d703a6e727":"performer name"
+}
+docker-compose up
+</code>