Faces: Add note on background filtering to background.go and docs photoprism#5167

Signed-off-by: Michael Mayer <[email protected]>

Author: lastzero

internal/ai/face/README.md

@@ -13,6 +13,8 @@ Key changes:
 - All face embeddings are now L2-normalized at creation, midpoint calculation, and deserialization time to keep cosine and Euclidean comparisons consistent.
 - Benchmarks were added to track the cost of hotspot routines (`Embedding.Dist` and `EmbeddingsMidpoint`).
 
+> **TODO:** Persist detector provenance in `FaceSrc` (e.g., use `entity.SrcONNX` for SCRFD detections) so hybrid libraries can toggle background filtering per embedding source when upgrading from Pigo.
+
 ### Detection Pipeline
 
 PhotoPrism now supports two interchangeable detection engines:

internal/ai/face/background.go

@@ -11,6 +11,18 @@ func (m Embedding) IsBackground() bool {
 
 // Background contains out-of-distribution (OOD) embeddings
 // of inputs that are not faces, such as plants, the sky, etc.
+//
+// NOTE: Checking an embedding against this list requires ~3–4 µs when it
+// fails to match (BenchmarkIsBackgroundMiss, temporary bench) because every
+// call scans all 24 clusters. We currently invoke IsBackground from:
+//   - Embedding.CanMatch() -> NewEmbeddings() at ingest time
+//   - Embeddings.Kind() / Embedding.SkipMatching() via entity.Face.SetEmbeddings()
+//   - buildFaceIndex() in internal/photoprism/faces_match.go
+//
+// Enabling IgnoreBackground therefore shortens index rebuilds roughly 5×,
+// but also removes the legacy guard that filtered Pigo false positives. When
+// we add detector provenance (TODO: record entity.SrcONNX on SCRFD detections),
+// we can conditionally skip this scan for ONNX-derived embeddings only.
 var Background = Clusters{
 	{
 		Radius:    0.35,