diff --git a/mobile/lib/face/db.dart b/mobile/lib/face/db.dart
index 6c4c27563..add547376 100644
--- a/mobile/lib/face/db.dart
+++ b/mobile/lib/face/db.dart
@@ -1,4 +1,5 @@
 import 'dart:async';
+import "dart:convert" show json;
 import "dart:io" show Directory;
 import "dart:math";
 
@@ -10,6 +11,7 @@ import 'package:path_provider/path_provider.dart';
 import "package:photos/extensions/stop_watch.dart";
 import 'package:photos/face/db_fields.dart';
 import "package:photos/face/db_model_mappers.dart";
+import "package:photos/face/model/detection.dart";
 import "package:photos/face/model/face.dart";
 import "package:photos/models/file/file.dart";
 import "package:photos/services/machine_learning/face_ml/face_clustering/face_info_for_clustering.dart";
@@ -455,6 +457,31 @@ class FaceMLDataDB {
     );
   }
 
+  // TODO: remove this method and replace by correct logic during indexing
+  Future<Map<String, bool>> getFaceIdsToIsSidewaysFaceTEMP() async {
+    final db = await instance.sqliteAsyncDB;
+    final List<Map<String, dynamic>> maps = await db.getAll(
+      'SELECT $faceIDColumn, $faceDetectionColumn FROM $facesTable',
+    );
+
+    final time = DateTime.now();
+    final Map<String, bool> result = {};
+    for (final map in maps) {
+      final faceID = map[faceIDColumn] as String;
+      final detection =
+          Detection.fromJson(json.decode(map[faceDetectionColumn] as String));
+      result[faceID] = detection.faceIsSideways();
+    }
+    _logger.info(
+      'decoding detections and calculating face sideways bools took ${DateTime.now().difference(time).inMilliseconds} ms',
+    );
+    final double sidewaysRatio =
+        result.values.where((e) => e).length / result.length;
+    _logger.info('sideways face ratio: $sidewaysRatio');
+
+    return result;
+  }
+
   Future<Set<FaceInfoForClustering>> getFaceInfoForClustering({
     double minScore = kMinHighQualityFaceScore,
     int minClarity = kLaplacianHardThreshold,
@@ -468,6 +495,9 @@ class FaceMLDataDB {
     );
     final db = await instance.sqliteAsyncDB;
 
+    final Map<String, bool> faceIdsToIsSideways =
+        await getFaceIdsToIsSidewaysFaceTEMP();
+
     final Set<FaceInfoForClustering> result = {};
     while (true) {
       // Query a batch of rows
@@ -494,6 +524,7 @@ class FaceMLDataDB {
           embeddingBytes: map[faceEmbeddingBlob] as Uint8List,
           faceScore: map[faceScore] as double,
           blurValue: map[faceBlur] as double,
+          isSideways: faceIdsToIsSideways[faceID] ?? false,
         );
         result.add(faceInfo);
       }
diff --git a/mobile/lib/face/model/detection.dart b/mobile/lib/face/model/detection.dart
index af6324ac8..49e8c3652 100644
--- a/mobile/lib/face/model/detection.dart
+++ b/mobile/lib/face/model/detection.dart
@@ -22,7 +22,7 @@ class Detection {
 
   bool get isEmpty => box.width == 0 && box.height == 0 && landmarks.isEmpty;
 
-  // emoty box
+  // empty box
   Detection.empty()
       : box = FaceBox(
           xMin: 0,
@@ -94,6 +94,9 @@ class Detection {
   }
 
   FaceDirection getFaceDirection() {
+    if (isEmpty) {
+      return FaceDirection.straight;
+    }
     final leftEye = [landmarks[0].x, landmarks[0].y];
     final rightEye = [landmarks[1].x, landmarks[1].y];
     final nose = [landmarks[2].x, landmarks[2].y];
@@ -131,6 +134,9 @@ class Detection {
   }
 
   bool faceIsSideways() {
+    if (isEmpty) {
+      return false;
+    }
     final leftEye = [landmarks[0].x, landmarks[0].y];
     final rightEye = [landmarks[1].x, landmarks[1].y];
     final nose = [landmarks[2].x, landmarks[2].y];
diff --git a/mobile/lib/services/machine_learning/face_ml/face_clustering/face_clustering_service.dart b/mobile/lib/services/machine_learning/face_ml/face_clustering/face_clustering_service.dart
index 7d822f9e7..792ec3b4c 100644
--- a/mobile/lib/services/machine_learning/face_ml/face_clustering/face_clustering_service.dart
+++ b/mobile/lib/services/machine_learning/face_ml/face_clustering/face_clustering_service.dart
@@ -452,7 +452,8 @@ class FaceClusteringService {
           faceScore: face.faceScore,
           blurValue: face.blurValue,
           badFace: face.faceScore < kMinHighQualityFaceScore ||
-              face.blurValue < kLaplacianSoftThreshold,
+              face.blurValue < kLaplacianSoftThreshold ||
+              face.isSideways,
           vEmbedding: Vector.fromList(
             EVector.fromBuffer(face.embeddingBytes).values,
             dtype: DType.float32,
@@ -606,7 +607,7 @@ class FaceClusteringService {
     );
     if (useDynamicThreshold) {
       log(
-        "[ClusterIsolate] ${DateTime.now()} Dynamic thresholding: $dynamicThresholdCount faces had a low face score or high blur value",
+        "[ClusterIsolate] ${DateTime.now()} Dynamic thresholding: $dynamicThresholdCount faces had a low face score or low blur clarity",
       );
     }
 
diff --git a/mobile/lib/services/machine_learning/face_ml/face_clustering/face_info_for_clustering.dart b/mobile/lib/services/machine_learning/face_ml/face_clustering/face_info_for_clustering.dart
index 8dd78ff1f..273d85da5 100644
--- a/mobile/lib/services/machine_learning/face_ml/face_clustering/face_info_for_clustering.dart
+++ b/mobile/lib/services/machine_learning/face_ml/face_clustering/face_info_for_clustering.dart
@@ -1,4 +1,3 @@
-
 import "dart:typed_data" show Uint8List;
 
 class FaceInfoForClustering {
@@ -7,6 +6,7 @@ class FaceInfoForClustering {
   final Uint8List embeddingBytes;
   final double faceScore;
   final double blurValue;
+  final bool isSideways;
 
   FaceInfoForClustering({
     required this.faceID,
@@ -14,5 +14,6 @@ class FaceInfoForClustering {
     required this.embeddingBytes,
     required this.faceScore,
     required this.blurValue,
+    this.isSideways = false,
   });
-}
\ No newline at end of file
+}