Merge remote-tracking branch 'origin/mobile_face' into mobile_face

mobile/lib/face/db.dart

@@ -78,6 +78,37 @@ class FaceMLDataDB {
     }
   }
 
+  Future<void> updatePersonIDForFaceIDIFNotSet(
+    Map<String, int> faceIDToPersonID,
+  ) async {
+    final db = await instance.database;
+    const batchSize = 500;
+    final numBatches = (faceIDToPersonID.length / batchSize).ceil();
+
+    for (int i = 0; i < numBatches; i++) {
+      _logger.info('updatePersonIDForFaceIDIFNotSet Batch $i of $numBatches');
+      final start = i * batchSize;
+      final end = min((i + 1) * batchSize, faceIDToPersonID.length);
+      final batch = faceIDToPersonID.entries.toList().sublist(start, end);
+
+      final batchUpdate = db.batch();
+
+      for (final entry in batch) {
+        final faceID = entry.key;
+        final personID = entry.value;
+
+        batchUpdate.update(
+          facesTable,
+          {faceClusterId: personID},
+          where: '$faceIDColumn = ? AND $faceClusterId IS NULL',
+          whereArgs: [faceID],
+        );
+      }
+
+      await batchUpdate.commit(noResult: true);
+    }
+  }
+
   Future<Map<int, int>> getIndexedFileIds() async {
     final db = await instance.database;
     final List<Map<String, dynamic>> maps = await db.rawQuery(
@@ -293,28 +324,6 @@ class FaceMLDataDB {
     return result;
   }
 
-  Future<void> updatePersonIDForFaceIDIFNotSet(
-    Map<String, int> faceIDToPersonID,
-  ) async {
-    final db = await instance.database;
-
-    // Start a batch
-    final batch = db.batch();
-
-    for (final map in faceIDToPersonID.entries) {
-      final faceID = map.key;
-      final personID = map.value;
-      batch.update(
-        facesTable,
-        {faceClusterId: personID},
-        where: '$faceIDColumn = ? AND $faceClusterId IS NULL',
-        whereArgs: [faceID],
-      );
-    }
-    // Commit the batch
-    await batch.commit(noResult: true);
-  }
-
   Future<void> forceUpdateClusterIds(
     Map<String, int> faceIDToPersonID,
   ) async {
@@ -347,7 +356,9 @@ class FaceMLDataDB {
     int offset = 0,
     int batchSize = 10000,
   }) async {
-    _logger.info('reading as float');
+    _logger.info(
+      'reading as float offset: $offset, maxFaces: $maxFaces, batchSize: $batchSize',
+    );
     final db = await instance.database;
 
     final Map<String, (int?, Uint8List)> result = {};

mobile/lib/services/machine_learning/face_ml/face_ml_service.dart

@@ -17,6 +17,7 @@ import "package:photos/db/files_db.dart";
 import "package:photos/db/ml_data_db.dart";
 import "package:photos/events/diff_sync_complete_event.dart";
 import "package:photos/extensions/list.dart";
+import "package:photos/extensions/stop_watch.dart";
 import "package:photos/face/db.dart";
 import "package:photos/face/model/box.dart";
 import "package:photos/face/model/detection.dart" as face_detection;
@@ -120,7 +121,12 @@ class FaceMlService {
       if (LocalSettings.instance.isFaceIndexingEnabled == false) {
         return;
       }
-      unawaited(indexAllImages());
+      // [neeraj] intentional delay in starting indexing on diff sync, this gives time for the user
+      // to disable face-indexing in case it's causing crash. In the future, we
+      // should have a better way to handle this.
+      Future.delayed(const Duration(seconds: 10), () {
+        unawaited(indexAllImages());
+      });
     });
   }
 
@@ -380,8 +386,8 @@ class FaceMlService {
             await FilesDB.instance.getFileIDToCreationTime();
 
         const int bucketSize = 10000;
+        const int batchSize = 10000;
         const int offsetIncrement = 7500;
-        const int batchSize = 5000;
         int offset = 0;
 
         while (true) {
@@ -413,8 +419,11 @@ class FaceMlService {
 
           await FaceMLDataDB.instance
               .updatePersonIDForFaceIDIFNotSet(faceIdToCluster);
-
-          offset += offsetIncrement;
+          if (offset == 0) {
+            offset += offsetIncrement;
+          } else {
+            offset += bucketSize;
+          }
         }
       } else {
         // Read all the embeddings from the database, in a map from faceID to embedding
@@ -518,8 +527,11 @@ class FaceMlService {
           fileIds.add(f.uploadedFileID!);
         }
         try {
+          final EnteWatch? w = kDebugMode ? EnteWatch("face_em_fetch") : null;
+          w?.start();
           final res =
               await RemoteFileMLService.instance.getFilessEmbedding(fileIds);
+          w?.logAndReset('fetched ${res.mlData.length} embeddings');
           final List<Face> faces = [];
           final remoteFileIdToVersion = <int, int>{};
           for (FileMl fileMl in res.mlData.values) {
@@ -537,6 +549,7 @@ class FaceMlService {
             remoteFileIdToVersion[fileMl.fileID] = fileMl.faceEmbedding.version;
           }
           await FaceMLDataDB.instance.bulkInsertFaces(faces);
+          w?.logAndReset('stored embeddings');
           for (final entry in remoteFileIdToVersion.entries) {
             alreadyIndexedFiles[entry.key] = entry.value;
           }
@@ -1242,7 +1255,6 @@ class FaceMlService {
     if (!enteFile.isUploaded || enteFile.isOwner == false) {
       return true;
     }
-
     // Skip if the file is a video
     if (enteFile.fileType == FileType.video) {
       return true;

mobile/lib/services/machine_learning/face_ml/feedback/cluster_feedback.dart

@@ -79,7 +79,7 @@ class ClusterFeedbackService {
     final personClusters = await faceMlDb.getPersonClusterIDs(p.remoteID);
     dev.log(
       'existing clusters for ${p.attr.name} are $personClusters',
-      name: "ClusterFeedbackService",
+      name: "getSuggestionsUsingMedian",
     );
 
     // Get and update the cluster summary to get the avg (centroid) and count
@@ -353,6 +353,9 @@ class ClusterFeedbackService {
     Set<int> ignoredClusters,
   ) async {
     final faceMlDb = FaceMLDataDB.instance;
+    _logger.info(
+      'start getUpdateClusterAvg for ${allClusterIdsToCountMap.length} clusters',
+    );
 
     final Map<int, (Uint8List, int)> clusterToSummary =
         await faceMlDb.clusterSummaryAll();
@@ -389,6 +392,7 @@ class ClusterFeedbackService {
     if (updatesForClusterSummary.isNotEmpty) {
       await faceMlDb.clusterSummaryUpdate(updatesForClusterSummary);
     }
+    _logger.info('end getUpdateClusterAvg for ${clusterAvg.length} clusters');
 
     return clusterAvg;
   }

mobile/lib/ui/settings/debug/face_debug_section_widget.dart

@@ -158,7 +158,8 @@ class _FaceDebugSectionWidgetState extends State<FaceDebugSectionWidget> {
           trailingIcon: Icons.chevron_right_outlined,
           trailingIconIsMuted: true,
           onTap: () async {
-            await FaceMlService.instance.clusterAllImages(minFaceScore: 0.75);
+            await FaceMlService.instance
+                .clusterAllImages(minFaceScore: 0.75, clusterInBuckets: true);
             Bus.instance.fire(PeopleChangedEvent());
             showShortToast(context, "Done");
           },
@@ -205,7 +206,7 @@ class _FaceDebugSectionWidgetState extends State<FaceDebugSectionWidget> {
         if (kDebugMode)
           MenuItemWidget(
             captionedTextWidget: const CaptionedTextWidget(
-              title: "Pull Embeddings From Local",
+              title: "Compute suggestions",
             ),
             pressedColor: getEnteColorScheme(context).fillFaint,
             trailingIcon: Icons.chevron_right_outlined,

mobile/lib/ui/viewer/people/person_cluster_suggestion.dart

@@ -37,8 +37,6 @@ class _PersonClustersState extends State<PersonReviewClusterSuggestion> {
     super.initState();
     // Initialize the future in initState
     _fetchClusterSuggestions();
-    // futureClusterSuggestions = ClusterFeedbackService.instance
-    //     .getClusterFilesForPersonID(widget.person);
   }
 
   @override