mirror of
https://github.com/ente-io/ente.git
synced 2025-08-08 07:28:26 +00:00
[mob][photos] Cleanup feedback
This commit is contained in:
laurenspriem
parent
7c535adff1
commit
aab7e39ab3
@ -1080,16 +1080,4 @@ class FaceMLDataDB {
|
|||||||
}
|
}
|
||||||
await forceUpdateClusterIds(faceIDToClusterID);
|
await forceUpdateClusterIds(faceIDToClusterID);
|
||||||
}
|
}
|
||||||
|
|
||||||
Future<void> addFacesToCluster(
|
|
||||||
List<String> faceIDs,
|
|
||||||
int clusterID,
|
|
||||||
) async {
|
|
||||||
final faceIDToClusterID = <String, int>{};
|
|
||||||
for (final faceID in faceIDs) {
|
|
||||||
faceIDToClusterID[faceID] = clusterID;
|
|
||||||
}
|
|
||||||
|
|
||||||
await forceUpdateClusterIds(faceIDToClusterID);
|
|
||||||
}
|
|
||||||
}
|
}
|
||||||
|
|||||||
@ -244,8 +244,12 @@ class ClusterFeedbackService {
|
|||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
Future<void> addFilesToCluster(List<String> faceIDs, int clusterID) async {
|
Future<void> addFacesToCluster(List<String> faceIDs, int clusterID) async {
|
||||||
await FaceMLDataDB.instance.addFacesToCluster(faceIDs, clusterID);
|
final faceIDToClusterID = <String, int>{};
|
||||||
|
for (final faceID in faceIDs) {
|
||||||
|
faceIDToClusterID[faceID] = clusterID;
|
||||||
|
}
|
||||||
|
await FaceMLDataDB.instance.forceUpdateClusterIds(faceIDToClusterID);
|
||||||
Bus.instance.fire(PeopleChangedEvent());
|
Bus.instance.fire(PeopleChangedEvent());
|
||||||
return;
|
return;
|
||||||
}
|
}
|
||||||
@ -411,7 +415,6 @@ class ClusterFeedbackService {
|
|||||||
return susClusters;
|
return susClusters;
|
||||||
}
|
}
|
||||||
|
|
||||||
// TODO: iterate over this method to find sweet spot
|
|
||||||
Future<ClusteringResult> breakUpCluster(
|
Future<ClusteringResult> breakUpCluster(
|
||||||
int clusterID, {
|
int clusterID, {
|
||||||
bool useDbscan = false,
|
bool useDbscan = false,
|
||||||
@ -468,178 +471,6 @@ class ClusterFeedbackService {
|
|||||||
return clusterResult;
|
return clusterResult;
|
||||||
}
|
}
|
||||||
|
|
||||||
/// WARNING: this method is purely for debugging purposes, never use in production
|
|
||||||
Future<void> createFakeClustersByBlurValue() async {
|
|
||||||
try {
|
|
||||||
// Delete old clusters
|
|
||||||
await FaceMLDataDB.instance.dropClustersAndPersonTable();
|
|
||||||
final List<PersonEntity> persons =
|
|
||||||
await PersonService.instance.getPersons();
|
|
||||||
for (final PersonEntity p in persons) {
|
|
||||||
await PersonService.instance.deletePerson(p.remoteID);
|
|
||||||
}
|
|
||||||
|
|
||||||
// Create new fake clusters based on blur value. One for values between 0 and 10, one for 10-20, etc till 200
|
|
||||||
final int startClusterID = DateTime.now().microsecondsSinceEpoch;
|
|
||||||
final faceIDsToBlurValues =
|
|
||||||
await FaceMLDataDB.instance.getFaceIDsToBlurValues(200);
|
|
||||||
final faceIdToCluster = <String, int>{};
|
|
||||||
for (final entry in faceIDsToBlurValues.entries) {
|
|
||||||
final faceID = entry.key;
|
|
||||||
final blurValue = entry.value;
|
|
||||||
final newClusterID = startClusterID + blurValue ~/ 10;
|
|
||||||
faceIdToCluster[faceID] = newClusterID;
|
|
||||||
}
|
|
||||||
await FaceMLDataDB.instance.updateFaceIdToClusterId(faceIdToCluster);
|
|
||||||
|
|
||||||
Bus.instance.fire(PeopleChangedEvent());
|
|
||||||
} catch (e, s) {
|
|
||||||
_logger.severe("Error in createFakeClustersByBlurValue", e, s);
|
|
||||||
rethrow;
|
|
||||||
}
|
|
||||||
}
|
|
||||||
|
|
||||||
Future<void> debugLogClusterBlurValues(
|
|
||||||
int clusterID, {
|
|
||||||
int? clusterSize,
|
|
||||||
bool logClusterSummary = false,
|
|
||||||
bool logBlurValues = false,
|
|
||||||
}) async {
|
|
||||||
if (!kDebugMode) return;
|
|
||||||
|
|
||||||
// Logging the clusterID
|
|
||||||
_logger.info(
|
|
||||||
"Debug logging for cluster $clusterID${clusterSize != null ? ' with $clusterSize photos' : ''}",
|
|
||||||
);
|
|
||||||
const int biggestClusterID = 1715061228725148;
|
|
||||||
|
|
||||||
// Logging the cluster summary for the cluster
|
|
||||||
if (logClusterSummary) {
|
|
||||||
final summaryMap = await FaceMLDataDB.instance.getClusterToClusterSummary(
|
|
||||||
[clusterID, biggestClusterID],
|
|
||||||
);
|
|
||||||
final summary = summaryMap[clusterID];
|
|
||||||
if (summary != null) {
|
|
||||||
_logger.info(
|
|
||||||
"Cluster summary for cluster $clusterID says the amount of faces is: ${summary.$2}",
|
|
||||||
);
|
|
||||||
}
|
|
||||||
|
|
||||||
final biggestClusterSummary = summaryMap[biggestClusterID];
|
|
||||||
final clusterSummary = summaryMap[clusterID];
|
|
||||||
if (biggestClusterSummary != null && clusterSummary != null) {
|
|
||||||
_logger.info(
|
|
||||||
"Cluster summary for biggest cluster $biggestClusterID says the size is: ${biggestClusterSummary.$2}",
|
|
||||||
);
|
|
||||||
_logger.info(
|
|
||||||
"Cluster summary for current cluster $clusterID says the size is: ${clusterSummary.$2}",
|
|
||||||
);
|
|
||||||
|
|
||||||
// Mean distance
|
|
||||||
final biggestMean = Vector.fromList(
|
|
||||||
EVector.fromBuffer(biggestClusterSummary.$1).values,
|
|
||||||
dtype: DType.float32,
|
|
||||||
);
|
|
||||||
final currentMean = Vector.fromList(
|
|
||||||
EVector.fromBuffer(clusterSummary.$1).values,
|
|
||||||
dtype: DType.float32,
|
|
||||||
);
|
|
||||||
final bigClustersMeanDistance = 1 - biggestMean.dot(currentMean);
|
|
||||||
_logger.info(
|
|
||||||
"Mean distance between biggest cluster and current cluster: $bigClustersMeanDistance",
|
|
||||||
);
|
|
||||||
_logger.info(
|
|
||||||
'Element differences between the two means are ${biggestMean - currentMean}',
|
|
||||||
);
|
|
||||||
final currentL2Norm = currentMean.norm();
|
|
||||||
_logger.info(
|
|
||||||
'L2 norm of current mean: $currentL2Norm',
|
|
||||||
);
|
|
||||||
final trueDistance =
|
|
||||||
biggestMean.distanceTo(currentMean, distance: Distance.cosine);
|
|
||||||
_logger.info('True distance between the two means: $trueDistance');
|
|
||||||
|
|
||||||
// Median distance
|
|
||||||
const sampleSize = 100;
|
|
||||||
final Iterable<Uint8List> biggestEmbeddings = await FaceMLDataDB
|
|
||||||
.instance
|
|
||||||
.getFaceEmbeddingsForCluster(biggestClusterID);
|
|
||||||
final List<Uint8List> biggestSampledEmbeddingsProto =
|
|
||||||
_randomSampleWithoutReplacement(
|
|
||||||
biggestEmbeddings,
|
|
||||||
sampleSize,
|
|
||||||
);
|
|
||||||
final List<Vector> biggestSampledEmbeddings =
|
|
||||||
biggestSampledEmbeddingsProto
|
|
||||||
.map(
|
|
||||||
(embedding) => Vector.fromList(
|
|
||||||
EVector.fromBuffer(embedding).values,
|
|
||||||
dtype: DType.float32,
|
|
||||||
),
|
|
||||||
)
|
|
||||||
.toList(growable: false);
|
|
||||||
|
|
||||||
final Iterable<Uint8List> currentEmbeddings =
|
|
||||||
await FaceMLDataDB.instance.getFaceEmbeddingsForCluster(clusterID);
|
|
||||||
final List<Uint8List> currentSampledEmbeddingsProto =
|
|
||||||
_randomSampleWithoutReplacement(
|
|
||||||
currentEmbeddings,
|
|
||||||
sampleSize,
|
|
||||||
);
|
|
||||||
final List<Vector> currentSampledEmbeddings =
|
|
||||||
currentSampledEmbeddingsProto
|
|
||||||
.map(
|
|
||||||
(embedding) => Vector.fromList(
|
|
||||||
EVector.fromBuffer(embedding).values,
|
|
||||||
dtype: DType.float32,
|
|
||||||
),
|
|
||||||
)
|
|
||||||
.toList(growable: false);
|
|
||||||
|
|
||||||
// Calculate distances and find the median
|
|
||||||
final List<double> distances = [];
|
|
||||||
final List<double> trueDistances = [];
|
|
||||||
for (final biggestEmbedding in biggestSampledEmbeddings) {
|
|
||||||
for (final currentEmbedding in currentSampledEmbeddings) {
|
|
||||||
distances.add(1 - biggestEmbedding.dot(currentEmbedding));
|
|
||||||
trueDistances.add(
|
|
||||||
biggestEmbedding.distanceTo(
|
|
||||||
currentEmbedding,
|
|
||||||
distance: Distance.cosine,
|
|
||||||
),
|
|
||||||
);
|
|
||||||
}
|
|
||||||
}
|
|
||||||
distances.sort();
|
|
||||||
trueDistances.sort();
|
|
||||||
final double medianDistance = distances[distances.length ~/ 2];
|
|
||||||
final double trueMedianDistance =
|
|
||||||
trueDistances[trueDistances.length ~/ 2];
|
|
||||||
_logger.info(
|
|
||||||
"Median distance between biggest cluster and current cluster: $medianDistance (using sample of $sampleSize)",
|
|
||||||
);
|
|
||||||
_logger.info(
|
|
||||||
'True distance median between the two embeddings: $trueMedianDistance',
|
|
||||||
);
|
|
||||||
}
|
|
||||||
}
|
|
||||||
|
|
||||||
// Logging the blur values for the cluster
|
|
||||||
if (logBlurValues) {
|
|
||||||
final List<double> blurValues = await FaceMLDataDB.instance
|
|
||||||
.getBlurValuesForCluster(clusterID)
|
|
||||||
.then((value) => value.toList());
|
|
||||||
final blurValuesIntegers =
|
|
||||||
blurValues.map((value) => value.round()).toList();
|
|
||||||
blurValuesIntegers.sort();
|
|
||||||
_logger.info(
|
|
||||||
"Blur values for cluster $clusterID${clusterSize != null ? ' with $clusterSize photos' : ''}: $blurValuesIntegers",
|
|
||||||
);
|
|
||||||
}
|
|
||||||
|
|
||||||
return;
|
|
||||||
}
|
|
||||||
|
|
||||||
/// Returns a list of suggestions. For each suggestion we return a record consisting of the following elements:
|
/// Returns a list of suggestions. For each suggestion we return a record consisting of the following elements:
|
||||||
/// 1. clusterID: the ID of the cluster
|
/// 1. clusterID: the ID of the cluster
|
||||||
/// 2. distance: the distance between the person's cluster and the suggestion
|
/// 2. distance: the distance between the person's cluster and the suggestion
|
||||||
@ -1126,6 +957,147 @@ class ClusterFeedbackService {
|
|||||||
"Sorting suggestions based on distance to person took ${endTime.difference(startTime).inMilliseconds} ms for ${suggestions.length} suggestions, of which ${clusterSummaryCallTime.difference(startTime).inMilliseconds} ms was spent on the cluster summary call",
|
"Sorting suggestions based on distance to person took ${endTime.difference(startTime).inMilliseconds} ms for ${suggestions.length} suggestions, of which ${clusterSummaryCallTime.difference(startTime).inMilliseconds} ms was spent on the cluster summary call",
|
||||||
);
|
);
|
||||||
}
|
}
|
||||||
|
|
||||||
|
Future<void> debugLogClusterBlurValues(
|
||||||
|
int clusterID, {
|
||||||
|
int? clusterSize,
|
||||||
|
bool logClusterSummary = false,
|
||||||
|
bool logBlurValues = false,
|
||||||
|
}) async {
|
||||||
|
if (!kDebugMode) return;
|
||||||
|
|
||||||
|
// Logging the clusterID
|
||||||
|
_logger.info(
|
||||||
|
"Debug logging for cluster $clusterID${clusterSize != null ? ' with $clusterSize photos' : ''}",
|
||||||
|
);
|
||||||
|
const int biggestClusterID = 1715061228725148;
|
||||||
|
|
||||||
|
// Logging the cluster summary for the cluster
|
||||||
|
if (logClusterSummary) {
|
||||||
|
final summaryMap = await FaceMLDataDB.instance.getClusterToClusterSummary(
|
||||||
|
[clusterID, biggestClusterID],
|
||||||
|
);
|
||||||
|
final summary = summaryMap[clusterID];
|
||||||
|
if (summary != null) {
|
||||||
|
_logger.info(
|
||||||
|
"Cluster summary for cluster $clusterID says the amount of faces is: ${summary.$2}",
|
||||||
|
);
|
||||||
|
}
|
||||||
|
|
||||||
|
final biggestClusterSummary = summaryMap[biggestClusterID];
|
||||||
|
final clusterSummary = summaryMap[clusterID];
|
||||||
|
if (biggestClusterSummary != null && clusterSummary != null) {
|
||||||
|
_logger.info(
|
||||||
|
"Cluster summary for biggest cluster $biggestClusterID says the size is: ${biggestClusterSummary.$2}",
|
||||||
|
);
|
||||||
|
_logger.info(
|
||||||
|
"Cluster summary for current cluster $clusterID says the size is: ${clusterSummary.$2}",
|
||||||
|
);
|
||||||
|
|
||||||
|
// Mean distance
|
||||||
|
final biggestMean = Vector.fromList(
|
||||||
|
EVector.fromBuffer(biggestClusterSummary.$1).values,
|
||||||
|
dtype: DType.float32,
|
||||||
|
);
|
||||||
|
final currentMean = Vector.fromList(
|
||||||
|
EVector.fromBuffer(clusterSummary.$1).values,
|
||||||
|
dtype: DType.float32,
|
||||||
|
);
|
||||||
|
final bigClustersMeanDistance = 1 - biggestMean.dot(currentMean);
|
||||||
|
_logger.info(
|
||||||
|
"Mean distance between biggest cluster and current cluster: $bigClustersMeanDistance",
|
||||||
|
);
|
||||||
|
_logger.info(
|
||||||
|
'Element differences between the two means are ${biggestMean - currentMean}',
|
||||||
|
);
|
||||||
|
final currentL2Norm = currentMean.norm();
|
||||||
|
_logger.info(
|
||||||
|
'L2 norm of current mean: $currentL2Norm',
|
||||||
|
);
|
||||||
|
final trueDistance =
|
||||||
|
biggestMean.distanceTo(currentMean, distance: Distance.cosine);
|
||||||
|
_logger.info('True distance between the two means: $trueDistance');
|
||||||
|
|
||||||
|
// Median distance
|
||||||
|
const sampleSize = 100;
|
||||||
|
final Iterable<Uint8List> biggestEmbeddings = await FaceMLDataDB
|
||||||
|
.instance
|
||||||
|
.getFaceEmbeddingsForCluster(biggestClusterID);
|
||||||
|
final List<Uint8List> biggestSampledEmbeddingsProto =
|
||||||
|
_randomSampleWithoutReplacement(
|
||||||
|
biggestEmbeddings,
|
||||||
|
sampleSize,
|
||||||
|
);
|
||||||
|
final List<Vector> biggestSampledEmbeddings =
|
||||||
|
biggestSampledEmbeddingsProto
|
||||||
|
.map(
|
||||||
|
(embedding) => Vector.fromList(
|
||||||
|
EVector.fromBuffer(embedding).values,
|
||||||
|
dtype: DType.float32,
|
||||||
|
),
|
||||||
|
)
|
||||||
|
.toList(growable: false);
|
||||||
|
|
||||||
|
final Iterable<Uint8List> currentEmbeddings =
|
||||||
|
await FaceMLDataDB.instance.getFaceEmbeddingsForCluster(clusterID);
|
||||||
|
final List<Uint8List> currentSampledEmbeddingsProto =
|
||||||
|
_randomSampleWithoutReplacement(
|
||||||
|
currentEmbeddings,
|
||||||
|
sampleSize,
|
||||||
|
);
|
||||||
|
final List<Vector> currentSampledEmbeddings =
|
||||||
|
currentSampledEmbeddingsProto
|
||||||
|
.map(
|
||||||
|
(embedding) => Vector.fromList(
|
||||||
|
EVector.fromBuffer(embedding).values,
|
||||||
|
dtype: DType.float32,
|
||||||
|
),
|
||||||
|
)
|
||||||
|
.toList(growable: false);
|
||||||
|
|
||||||
|
// Calculate distances and find the median
|
||||||
|
final List<double> distances = [];
|
||||||
|
final List<double> trueDistances = [];
|
||||||
|
for (final biggestEmbedding in biggestSampledEmbeddings) {
|
||||||
|
for (final currentEmbedding in currentSampledEmbeddings) {
|
||||||
|
distances.add(1 - biggestEmbedding.dot(currentEmbedding));
|
||||||
|
trueDistances.add(
|
||||||
|
biggestEmbedding.distanceTo(
|
||||||
|
currentEmbedding,
|
||||||
|
distance: Distance.cosine,
|
||||||
|
),
|
||||||
|
);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
distances.sort();
|
||||||
|
trueDistances.sort();
|
||||||
|
final double medianDistance = distances[distances.length ~/ 2];
|
||||||
|
final double trueMedianDistance =
|
||||||
|
trueDistances[trueDistances.length ~/ 2];
|
||||||
|
_logger.info(
|
||||||
|
"Median distance between biggest cluster and current cluster: $medianDistance (using sample of $sampleSize)",
|
||||||
|
);
|
||||||
|
_logger.info(
|
||||||
|
'True distance median between the two embeddings: $trueMedianDistance',
|
||||||
|
);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
// Logging the blur values for the cluster
|
||||||
|
if (logBlurValues) {
|
||||||
|
final List<double> blurValues = await FaceMLDataDB.instance
|
||||||
|
.getBlurValuesForCluster(clusterID)
|
||||||
|
.then((value) => value.toList());
|
||||||
|
final blurValuesIntegers =
|
||||||
|
blurValues.map((value) => value.round()).toList();
|
||||||
|
blurValuesIntegers.sort();
|
||||||
|
_logger.info(
|
||||||
|
"Blur values for cluster $clusterID${clusterSize != null ? ' with $clusterSize photos' : ''}: $blurValuesIntegers",
|
||||||
|
);
|
||||||
|
}
|
||||||
|
|
||||||
|
return;
|
||||||
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
/// Returns a map of person's clusterID to map of closest clusterID to with disstance
|
/// Returns a map of person's clusterID to map of closest clusterID to with disstance
|
||||||
|
|||||||
@ -279,7 +279,7 @@ class _FaceWidgetState extends State<FaceWidget> {
|
|||||||
try {
|
try {
|
||||||
if (isJustRemoved) {
|
if (isJustRemoved) {
|
||||||
await ClusterFeedbackService.instance
|
await ClusterFeedbackService.instance
|
||||||
.addFilesToCluster([widget.face.faceID], widget.clusterID!);
|
.addFacesToCluster([widget.face.faceID], widget.clusterID!);
|
||||||
} else {
|
} else {
|
||||||
await ClusterFeedbackService.instance
|
await ClusterFeedbackService.instance
|
||||||
.removeFilesFromCluster([widget.file], widget.clusterID!);
|
.removeFilesFromCluster([widget.file], widget.clusterID!);
|
||||||
|
|||||||
Loading…
x
Reference in New Issue
Block a user