-
-
Notifications
You must be signed in to change notification settings - Fork 47
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Add the sixth model to the application. Aka DlibTaguchiHog model. =)
In this model 6, we use the model newly trained by Tanguchi to obtain the descriptor and perform the face recognition. This model was trained from scratch to slightly improve the bias of the original model on non-Caucasian/American people. It obtained a similar result in the LFW tests, slightly lower, but within the acceptable margins of error. It may still have bias problems, they recognize that it fails with African people and with childrens, but in principle it seems like an improvement to the model used in 1, 2, 3 and 4. 😉 See: https://github.com/TaguchiModels/dlibModels This model is added for testing, and I understand that it will be of great help to those who have had problems with the original model.
- Loading branch information
1 parent
757eebb
commit 2beb9bd
Showing
5 changed files
with
155 additions
and
9 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,110 @@ | ||
| <?php | ||
| /** | ||
| * @copyright Copyright (c) 2020-2024, Matias De lellis <[email protected]> | ||
| * | ||
| * @author Matias De lellis <[email protected]> | ||
| * | ||
| * @license GNU AGPL version 3 or any later version | ||
| * | ||
| * This program is free software: you can redistribute it and/or modify | ||
| * it under the terms of the GNU Affero General Public License as | ||
| * published by the Free Software Foundation, either version 3 of the | ||
| * License, or (at your option) any later version. | ||
| * | ||
| * 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 | ||
| * along with this program. If not, see <http://www.gnu.org/licenses/>. | ||
| * | ||
| */ | ||
|
|
||
| namespace OCA\FaceRecognition\Model\DlibTaguchiHogModel; | ||
|
|
||
| use OCA\FaceRecognition\Helper\FaceRect; | ||
|
|
||
| use OCA\FaceRecognition\Model\DlibCnnModel\DlibCnnModel; | ||
| use OCA\FaceRecognition\Model\IModel; | ||
|
|
||
| class DlibTaguchiHogModel extends DlibCnnModel implements IModel { | ||
|
|
||
| /* | ||
| * Model files. | ||
| */ | ||
| const FACE_MODEL_ID = 6; | ||
| const FACE_MODEL_NAME = "DlibTaguchiHog"; | ||
| const FACE_MODEL_DESC = "Extends the Taguchi model, doing a face validation with the Hog detector"; | ||
| const FACE_MODEL_DOC = "https://github.com/matiasdelellis/facerecognition/wiki/Models#model-6"; | ||
|
|
||
| /** Relationship between image size and memory consumed */ | ||
| const MEMORY_AREA_RELATIONSHIP = 1 * 1024; | ||
| const MINIMUM_MEMORY_REQUIREMENTS = 1 * 1024 * 1024 * 1024; | ||
|
|
||
| /* | ||
| * Model files. | ||
| */ | ||
| const FACE_MODEL_FILES = [ | ||
| 'detector' => [ | ||
| 'url' => 'https://github.com/davisking/dlib-models/raw/94cdb1e40b1c29c0bfcaf7355614bfe6da19460e/mmod_human_face_detector.dat.bz2', | ||
| 'filename' => 'mmod_human_face_detector.dat' | ||
| ], | ||
| 'predictor' => [ | ||
| 'url' => 'https://github.com/davisking/dlib-models/raw/4af9b776281dd7d6e2e30d4a2d40458b1e254e40/shape_predictor_5_face_landmarks.dat.bz2', | ||
| 'filename' => 'shape_predictor_5_face_landmarks.dat', | ||
| ], | ||
| 'resnet' => [ | ||
| 'url' => 'https://github.com/TaguchiModels/dlibModels/raw/main/taguchi_face_recognition_resnet_model_v1.7z', | ||
| 'filename' => 'taguchi_face_recognition_resnet_model_v1.dat' | ||
| ] | ||
| ]; | ||
|
|
||
| public function detectFaces(string $imagePath, bool $compute = true): array { | ||
| $detectedFaces = []; | ||
|
|
||
| $cnnFaces = parent::detectFaces($imagePath); | ||
| if (count($cnnFaces) === 0) { | ||
| return $detectedFaces; | ||
| } | ||
|
|
||
| $hogFaces = dlib_face_detection($imagePath); | ||
|
|
||
| foreach ($cnnFaces as $proposedFace) { | ||
| $detectedFaces[] = $this->validateFace($proposedFace, $hogFaces); | ||
| } | ||
|
|
||
| return $detectedFaces; | ||
| } | ||
|
|
||
| private function validateFace($proposedFace, array $validateFaces) { | ||
| foreach ($validateFaces as $validateFace) { | ||
| $overlapPercent = FaceRect::overlapPercent($proposedFace, $validateFace); | ||
| /** | ||
| * The weak link in our default model is the landmark detector that | ||
| * can't align profile or rotate faces correctly. | ||
| * | ||
| * The Hog detector also fails and cannot detect these faces. So, we | ||
| * consider if Hog detector can detect it, to infer when the predictor | ||
| * will give good results. | ||
| * | ||
| * If Hog detects it (Overlap > 35%), we can assume that landmark | ||
| * detector will do it too. In this case, we consider the face valid, | ||
| * and just return it. | ||
| */ | ||
| if ($overlapPercent >= 0.35) { | ||
| return $proposedFace; | ||
| } | ||
| } | ||
|
|
||
| /** | ||
| * If Hog don't detect this face, they are probably in profile or rotated. | ||
| * These are bad to compare, so we lower the confidence, to avoid clustering. | ||
| */ | ||
| $confidence = $proposedFace['detection_confidence']; | ||
| $proposedFace['detection_confidence'] = $confidence * 0.6; | ||
|
|
||
| return $proposedFace; | ||
| } | ||
|
|
||
| } |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters