1e830a3285
This change adds support to retrieve 128D face descriptor for a given landmark. Since now we have full pipeline, README.md has "general usage" section and integration test is added. Also, return from FaceLandmarkDetection is changed, so it can be given to FaceRecognition without changes. All obtained values are crosschecked to match with values from python versions (however, if num_jitters is > 1 in FaceRecognition, values don't match between PHP and Python, I suspect it is related to usage of dlib::rand, but still investigating)..
60 lines
2.6 KiB
PHP
60 lines
2.6 KiB
PHP
--TEST--
|
|
Full test for face recognition - download models, detect faces, landmark detection and face recognition.
|
|
--SKIPIF--
|
|
<?php if (!extension_loaded("pdlib") || (function_exists("bzopen"))) print "skip"; ?>
|
|
--FILE--
|
|
<?php
|
|
$models = array(
|
|
"detection" => array("uri"=>"http://dlib.net/files/mmod_human_face_detector.dat.bz2"),
|
|
"prediction" => array("uri"=>"http://dlib.net/files/shape_predictor_5_face_landmarks.dat.bz2"),
|
|
"recognition" => array("uri"=>"http://dlib.net/files/dlib_face_recognition_resnet_model_v1.dat.bz2")
|
|
);
|
|
|
|
// Check if there are models in local tmp. Download them if not (lazy caching).
|
|
//
|
|
foreach ($models as $modelName => $modelBag) {
|
|
printf("Processing %s model\n", $modelName);
|
|
$bz2_filename = array_values(array_slice(explode("/", $modelBag["uri"]), -1))[0];
|
|
$temp_bz2_file = sys_get_temp_dir() . "/" . $bz2_filename;
|
|
$dat_filename = array_values(array_slice(explode(".", $bz2_filename), 0))[0] . ".dat";
|
|
$temp_dat_file = sys_get_temp_dir() . "/" . $dat_filename;
|
|
$models[$modelName]["local_path"] = $temp_dat_file;
|
|
|
|
if (file_exists($temp_dat_file)) {
|
|
continue;
|
|
}
|
|
file_put_contents($temp_bz2_file, fopen($modelBag["uri"], 'r'));
|
|
$bz = bzopen($temp_bz2_file, "r");
|
|
$decompressed_file = "";
|
|
while (!feof($bz)) {
|
|
$decompressed_file .= bzread($bz, 4096);
|
|
}
|
|
bzclose($bz);
|
|
|
|
file_put_contents($temp_dat_file, $decompressed_file);
|
|
}
|
|
|
|
printf("Detection\n");
|
|
$fd = new CnnFaceDetection($models["detection"]["local_path"]);
|
|
$detected_faces = $fd->detect(__DIR__ . "/lenna.jpg");
|
|
printf("Faces found = %d\n", count($detected_faces));
|
|
foreach($detected_faces as $index => $detected_face) {
|
|
printf("Face[%d] in bounding box (left=%d, top=%d, right=%d, bottom=%d)\n", $index,
|
|
$detected_face["left"], $detected_face["top"], $detected_face["right"], $detected_face["bottom"]);
|
|
$fld = new FaceLandmarkDetection($models["prediction"]["local_path"]);
|
|
$landmarks = $fld->detect(__DIR__ . "/lenna.jpg", $detected_face);
|
|
printf("Since we used model with 5 shape predictions, we found %d landmark parts\n", count($landmarks["parts"]));
|
|
$fr = new FaceRecognition($models["recognition"]["local_path"]);
|
|
$descriptor = $fr->computeDescriptor(__DIR__ . "/lenna.jpg", $landmarks);
|
|
printf("Descriptor is vector of %d dimensions\n", count($descriptor));
|
|
}
|
|
?>
|
|
--EXPECT--
|
|
Processing detection model
|
|
Processing prediction model
|
|
Processing recognition model
|
|
Detection
|
|
Faces found = 1
|
|
Face[0] in bounding box (left=187, top=186, right=357, bottom=355)
|
|
Since we used model with 5 shape predictions, we found 5 landmark parts
|
|
Descriptor is vector of 128 dimensions
|