Implement the Chinese Whispers clustering algorithm in PHP code.

.github/workflows/coverage.yml

@@ -15,7 +15,7 @@ jobs:
       matrix:
         php-versions: ['8.1']
         databases: ['mysql']
-        server-versions: ['stable26']
+        server-versions: ['stable27']
 
     name: php${{ matrix.php-versions }}-${{ matrix.databases }}-${{ matrix.server-versions }}
 

.github/workflows/phpunit.yml

@@ -19,7 +19,7 @@ jobs:
       matrix:
         php-versions: ['8.0', '8.1', '8.2']
         databases: ['sqlite', 'mysql', 'pgsql']
-        server-versions: ['stable26']
+        server-versions: ['stable27']
 
     name: php${{ matrix.php-versions }}-${{ matrix.databases }}-${{ matrix.server-versions }}
 

.github/workflows/static-analysis.yml

@@ -11,7 +11,7 @@ jobs:
       runs-on: ubuntu-latest
       strategy:
           matrix:
-              ocp-version: ['dev-stable26']
+              ocp-version: ['dev-stable27']
       name: Nextcloud ${{ matrix.ocp-version }}
       steps:
           - name: Checkout

appinfo/info.xml

@@ -16,7 +16,7 @@
 - **🚀 Build your own thing:** FaceRecognition app is just a basic building block. Through FaceRecognition API, you can build your advanced scenarios - automatically add tags to images, connect contacts and persons, share images from specific person… We want to hear your ideas!
 ]]>
 	</description>
-	<version>0.9.20</version>
+	<version>0.9.30</version>
 	<licence>agpl</licence>
 	<author>Matias De lellis</author>
 	<author>Branko Kokanovic</author>
@@ -34,9 +34,7 @@
 	<screenshot>https://matiasdelellis.github.io/img/facerecognition/facerecognition-assign-initial-name.jpeg</screenshot>
 	<dependencies>
 		<php min-version="8.0" max-version="8.2" />
-		<lib>pdlib</lib>
-		<lib>bz2</lib>
-		<nextcloud min-version="26" max-version="27"/>
+		<nextcloud min-version="27" max-version="27"/>
 	</dependencies>
 	<repair-steps>
 		<uninstall>

lib/BackgroundJob/Tasks/CheckRequirementsTask.php

@@ -90,8 +90,10 @@ public function execute(FaceRecognitionContext $context) {
 		$phpMemory = MemoryLimits::getPhpMemory();
 		$this->logDebug("PHP Memory Limit: " . ($phpMemory > 0 ? $phpMemory : "Unknown"));
 
+		$this->logDebug("Clustering backend: " . (Requirements::pdlibLoaded() ? "pdlib" : "PHP (Not recommended."));
+
 		if ($this->imaginaryHelper->isEnabled()) {
-			$this->logDebug("Backend of images: Imaginary");
+			$this->logDebug("Image Backend: Imaginary");
 			$version = $this->imaginaryHelper->getVersion();
 			if ($version) {
 				$this->logDebug("Imaginary version: " . $version);
@@ -104,15 +106,7 @@ public function execute(FaceRecognitionContext $context) {
 				return false;
 			}
 		} else {
-			$this->logDebug("Backend of images: Imagick");
-		}
-
-		if (!Requirements::pdlibLoaded()) {
-			$error_message =
-				"The PDlib PHP extension is not loaded. Cannot continue without it." .
-				"Please read the documentation again about how to install the application: https://github.com/matiasdelellis/facerecognition/wiki/Installation";
-			$this->logInfo($error_message);
-			return false;
+			$this->logDebug("Image Backend: Imagick");
 		}
 
 		if (!Requirements::hasEnoughMemory()) {

lib/BackgroundJob/Tasks/CreateClustersTask.php

@@ -1,6 +1,6 @@
 <?php
 /**
- * @copyright Copyright (c) 2017-2020 Matias De lellis <mati86dl@gmail.com>
+ * @copyright Copyright (c) 2017-2023 Matias De lellis <mati86dl@gmail.com>
  * @copyright Copyright (c) 2018, Branko Kokanovic <branko@kokanovic.org>
  *
  * @author Branko Kokanovic <branko@kokanovic.org>
@@ -33,6 +33,9 @@
 use OCA\FaceRecognition\Db\PersonMapper;
 
 use OCA\FaceRecognition\Helper\Euclidean;
+use OCA\FaceRecognition\Helper\Requirements;
+
+use OCA\FaceRecognition\Clusterer\ChineseWhispers;
 
 use OCA\FaceRecognition\Service\SettingsService;
 /**
@@ -282,10 +285,9 @@ private function getNewClusters(array $faces): array {
 		// Clustering parameters
 		$sensitivity = $this->settingsService->getSensitivity();
 
-		// Create edges for chinese whispers
-		$edges = array();
-
-		if (version_compare(phpversion('pdlib'), '1.0.2', '>=')) {
+		if (Requirements::pdlibLoaded()) {
+			// Create edges (neighbors) for Chinese Whispers
+			$edges = array();
 			$faces_count = count($faces);
 			for ($i = 0; $i < $faces_count; $i++) {
 				$face1 = $faces[$i];
@@ -304,8 +306,14 @@ private function getNewClusters(array $faces): array {
 					}
 				}
 			}
+
+			// Given the edges get the list of labels (found clusters) for each face.
+			$newChineseClustersByIndex = dlib_chinese_whispers($edges);
 		} else {
+			// Create edges (neighbors) for Chinese Whispers
+			$edges = array();
 			$faces_count = count($faces);
+
 			for ($i = 0; $i < $faces_count; $i++) {
 				$face1 = $faces[$i];
 				if (!isset($face1->descriptor)) {
@@ -323,17 +331,27 @@ private function getNewClusters(array $faces): array {
 					}
 				}
 			}
+
+			// The clustering algorithm actually expects ordered lists.
+			$oedges = [];
+			ChineseWhispers::convert_unordered_to_ordered($edges, $oedges);
+			usort($oedges, function($a, $b) {
+				if ($a[0] === $b[0]) return $a[1] - $b[1];
+				return $a[0] - $b[0];
+			});
+
+			// Given the edges get the list of labels (found clusters) for each face.
+			$newChineseClustersByIndex = [];
+			ChineseWhispers::predict($oedges, $newChineseClustersByIndex);
 		}
 
-		$newChineseClustersByIndex = dlib_chinese_whispers($edges);
 		$newClusters = array();
 		for ($i = 0, $c = count($newChineseClustersByIndex); $i < $c; $i++) {
 			if (!isset($newClusters[$newChineseClustersByIndex[$i]])) {
 				$newClusters[$newChineseClustersByIndex[$i]] = array();
 			}
 			$newClusters[$newChineseClustersByIndex[$i]][] = $faces[$i]->id;
 		}
-
 		return $newClusters;
 	}
 

lib/Clusterer/ChineseWhispers.php

@@ -0,0 +1,159 @@
+<?php
+declare(strict_types=1);
+/**
+ * @copyright Copyright (c) 2023, Matias De lellis
+ *
+ * @author Matias De lellis <mati86dl@gmail.com>
+ *
+ * @license AGPL-3.0-or-later
+ *
+ * This code is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU Affero General Public License, version 3,
+ * as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU Affero General Public License for more details.
+ *
+ * You should have received a copy of the GNU Affero General Public License, version 3,
+ * along with this program. If not, see <http://www.gnu.org/licenses/>
+ *
+ */
+
+namespace OCA\FaceRecognition\Clusterer;
+
+
+/**
+ * This class implements the graph clustering algorithm described in the
+ * paper: Chinese Whispers - an Efficient Graph Clustering Algorithm and its
+ * Application to Natural Language Processing Problems by Chris Biemann.
+ *
+ * In particular, it tries to be a shameless copy of the original dlib
+ * implementation.
+ *  - https://github.com/davisking/dlib/blob/master/dlib/clustering/chinese_whispers.h
+ */
+class ChineseWhispers {
+
+	/**
+	 * Cluster the dataset by assigning a label to each sample.from the edges
+	 */
+	static public function predict(array &$edges, array &$labels, int $num_iterations = 100)
+	{
+		// To improve the stability of the clusters, we must
+		// iterate the neighbors in a pseudo-random way.
+		mt_srand(2023);
+
+		$labels = [];
+		if (count($edges) == 0)
+			return 0;
+
+		$neighbors = [];
+		self::find_neighbor_ranges($edges, $neighbors);
+
+		// Initialize the labels, each node gets a different label.
+		for ($i = 0; $i < count($neighbors); ++$i)
+			$labels[$i] = $i;
+
+		for ($iter = 0; $iter < count($neighbors)*$num_iterations; ++$iter)
+		{
+			// Pick a random node.
+			$idx = mt_rand()%count($neighbors);
+
+			// Count how many times each label happens amongst our neighbors.
+			$labels_to_counts = [];
+			$end = $neighbors[$idx][1];
+
+			for ($i = $neighbors[$idx][0]; $i != $end; ++$i)
+			{
+				$iLabelFirst = $edges[$i][1];
+				$iLabel = $labels[$iLabelFirst];
+				if (isset($labels_to_counts[$iLabel]))
+					$labels_to_counts[$iLabel]++;
+				else
+					$labels_to_counts[$iLabel] = 1;
+			}
+
+			// find the most common label
+			// std::map<unsigned long, double>::iterator i;
+			$best_score = PHP_INT_MIN;
+			$best_label = $labels[$idx];
+			foreach ($labels_to_counts as $key => $value)
+			{
+				if ($value > $best_score)
+				{
+					$best_score = $value;
+					$best_label = $key;
+				}
+			}
+
+			$labels[$idx] = $best_label;
+		}
+
+		// Remap the labels into a contiguous range.  First we find the
+		// mapping.
+		$label_remap = [];
+		for ($i = 0; $i < count($labels); ++$i)
+		{
+			$next_id = count($label_remap);
+			if (!isset($label_remap[$labels[$i]]))
+				$label_remap[$labels[$i]] = $next_id;
+		}
+		// now apply the mapping to all the labels.
+		for ($i = 0; $i < count($labels); ++$i)
+		{
+			$labels[$i] = $label_remap[$labels[$i]];
+		}
+
+		return count($label_remap);
+	}
+
+	static function find_neighbor_ranges (&$edges, &$neighbors) {
+		// setup neighbors so that [neighbors[i].first, neighbors[i].second) is the range
+		// within edges that contains all node i's edges.
+		$num_nodes = self::max_index_plus_one($edges);
+		for ($i = 0; $i < $num_nodes; ++$i) $neighbors[$i] = [0, 0];
+		$cur_node = 0;
+		$start_idx = 0;
+		for ($i = 0; $i < count($edges); ++$i)
+		{
+			if ($edges[$i][0] != $cur_node)
+			{
+				$neighbors[$cur_node] = [$start_idx, $i];
+				$start_idx = $i;
+				$cur_node = $edges[$i][0];
+			}
+		}
+		if (count($neighbors) !== 0)
+			$neighbors[$cur_node] = [$start_idx, count($edges)];
+	}
+
+	static function max_index_plus_one ($pairs): int {
+		if (count($pairs) === 0)
+		{
+			return 0;
+		}
+		else {
+			$max_idx = 0;
+			for ($i = 0; $i < count($pairs); ++$i)
+			{
+				if ($pairs[$i][0] > $max_idx)
+					$max_idx = $pairs[$i][0];
+				if ($pairs[$i][1] > $max_idx)
+					$max_idx = $pairs[$i][1];
+			}
+			return $max_idx + 1;
+		}
+	}
+
+	static function convert_unordered_to_ordered (&$edges, &$out_edges)
+	{
+		$out_edges = [];
+		for ($i = 0; $i < count($edges); ++$i)
+		{
+			$out_edges[] = [$edges[$i][0], $edges[$i][1]];
+			if ($edges[$i][0] != $edges[$i][1])
+				$out_edges[] = [$edges[$i][1], $edges[$i][0]];
+		}
+	}
+}

lib/Model/ExternalModel/ExternalModel.php

@@ -87,10 +87,6 @@ public function isInstalled(): bool {
 	}
 
 	public function meetDependencies(string &$error_message): bool {
-		if (!extension_loaded('pdlib')) {
-			$error_message = "The PDlib PHP extension is not loaded.";
-			return false;
-		}
 		if (is_null($this->settingsService->getExternalModelUrl())) {
 			$error_message = "You still need to configure the URL of the service running the model.";
 			return false;