 |
ROC SDK
2.4.0
Scalable Face Recognition Software
|
|
ROC SDK
2.4.0
Scalable Face Recognition Software
|
Construct templates from images.
This command line application corresponds to the roc_represent function in the Application Programming Interface. For representing videos from the command line see roc-track.
| Option | Meaning | Default |
|---|---|---|
-b | Treat input files as OpenBR galleries | Treat input files as individual images |
-c <int> | Rotate the image using roc_rotate | Don't rotate the image |
-d <float>,... | Specify manual landmarks | Don't specify manual landmarks |
-e | Run in Evaluation Mode | Don't run in evaluation mode |
-f <float> | Specify face detection False Detection Rate | Use ROC_SUGGESTED_FALSE_DETECTION_RATE |
-g | Estimate face analytics with ROC_ANALYTICS | Don't estimate face analytics |
-i | Ignore detections provided in OpenBR galleries | Don't ignore ground truth detections |
-j | Maximum thread count | CPU thread count |
-k <int> | Specify the Maximum Faces per image | 1 |
-l | Treat input files as newline-separated image lists | Treat input files as individual images |
-m <int> | Specify face detection Minimum Size in pixels | Use ROC_SUGGESTED_ABSOLUTE_MIN_SIZE |
-min-representation-size <int> | Specify face representation minimum size in pixels | 0 |
-n <float> | Specify face detection roc_adaptive_minimum_size | Use ROC_SUGGESTED_RELATIVE_MIN_SIZE |
-o | Estimate pitch, yaw and lips with ROC_PITCHYAW | Don't estimate pitch, yaw and lips |
-r | Treat input files as folders of images | Treat input files as individual images |
-t | Construct face thumbnail with ROC_THUMBNAIL | Don't construct face thumbnail |
-p | Estimate ICAO quality metrics with ROC_ICAO_METRICS | Don't estimate ICAO quality metrics |
-u | Run ROC_ICAO_BACKGROUND quality metric | Don't estimate ICAO uniform background metric |
-y | Abort on an image decoding error | Print a warning |
-z <int>,<int>,<float>,<float> | Specify roc_set_thumbnail_parameters | Use default thumbnail parameters |
--standard | Use ROC_STANDARD_REPRESENTATION with other representations | Use only the specified representation |
--deep | Use ROC_DEEP_REPRESENTATION face representation | Use ROC_STANDARD_REPRESENTATION |
--fast | Use ROC_FAST_REPRESENTATION face representation | Use ROC_STANDARD_REPRESENTATION |
--rotate-on-fte | Rotate images and retry if face detection fails | Don't rotate and retry |
--spoof | Static face liveness estimation with ROC_SPOOF | Don't estimate face liveness |
--detection-log-level | Set detection face count log level | Don't set face count log level |
--cuda | Use CUDA acceleration | Don't use CUDA acceleration |
--min-quality <float> | Minimum template face quality | Use roc_get_suggested_quality_threshold |
--no-min-quality | Disable quality filtering | Remove templates below quality threshold |
--represent-tattoos | Use ROC_TATTOO_REPRESENTATION | Use ROC_STANDARD_REPRESENTATION |
--detect-tattoos | Use ROC_TATTOO_DETECTION | Don't detect tattoos |
--archive <path> | Copy images to the specified roc_database | Don't copy images |
--frontal | Use ROC_FRONTAL_DETECTION | Use ROC_STANDARD_DETECTION |
--ignore-artwork | Use ROC_IGNORE_ARTWORK | Don't use ROC_IGNORE_ARTWORK |
--ignore-partial | Use ROC_IGNORE_PARTIAL | Don't use ROC_IGNORE_PARTIAL |
--opportunistic-batch-size <int> | Process images jointly with roc_set_opportunistic_batching | Process images independently |
--opportunistic-max-latency <int> | Maximum wait when accumulating a batch for processing | Wait up to one second |
--exif | Include roc_read_exif output in template metadata | Don't include Exif in template metadata |
--model-path | Path to dynamically loaded models | Don't load extra models |
--detect-text | Use ROC_TEXT_DETECTION | Don't detect text |
--detect-airplanes | Use ROC_AIRPLANE_DETECTION to detect airplanes | Don't detect airplanes |
--detect-bicycles | Use ROC_BICYCLE_DETECTION to detect bicycles | Don't detect bicycles |
--detect-boats | Use ROC_BOAT_DETECTION to detect boats | Don't detect boats |
--detect-buses | Use ROC_BUS_DETECTION to detect buses | Don't detect buses |
--detect-cars | Use ROC_CAR_DETECTION to detect cars | Don't detect cars |
--detect-guns | Use ROC_GUN_DETECTION to detect guns | Don't detect guns |
--detect-license-plates | Use ROC_LICENSE_PLATE_DETECTION to detect license plates | Don't detect license plates |
--detect-military-vehicles | Use ROC_MILITARY_VEHICLE_DETECTION to detect military vehicles | Don't detect military vehicles |
--detect-motorcycles | Use ROC_MOTORCYCLE_DETECTION to detect motorcycles | Don't detect motorcycles |
--detect-people | Use ROC_PERSON_DETECTION to detect people | Don't detect people |
--detect-trucks | Use ROC_TRUCK_DETECTION to detect trucks | Don't detect trucks |
--detect-all-objects | Use ROC_ALL_OBJECT_DETECTION to detect all objects | Don't detect objects |
--represent-objects | Use ROC_OBJECT_REPRESENTATION to represent objects for search | Don't represent objects |
--represent-color | Use ROC_COLOR_REPRESENTATION to represent an object's color(s) | Don't represent color |
--recognize-text | Use ROC_TEXT_REPRESENTATION | Don't recognize text |
--fingerprint | Use ROC_FINGERPRINT_REPRESENTATION | Don't use ROC_FINGERPRINT_REPRESENTATION |
--finger-id | Set a roc_finger_options identifier for the image | Set finger id to ROC_UNKNOWN_FINGER |
--resolution | Set a image resolution for fingerprint recognition. | 500 PPI |
--num-fingers | Set the number of fingers expected in an image. | 1 |
--keep-alive | Attempt to recover from failed network transmissions | Generate an error on a failed network transmission |
--roi <int>,<int>,<int>,<int> | Specify roc_represent_roi region of interest | Don't crop images |
--log-stdout> | Log messages to STDOUT | Don't log messages |
--log-file <file> | Log messages to a file | Don't log messages |
--sender-id | Transmission identifier unique to this application | roc_uuid_get_random |
--background-color <R>,<G>,<B> | Specify roc_set_icao_background_color | Use default background color |
When evaluating the accuracy of an algorithm it is often helpful to enforce the count and order of the templates. In evaluation mode exactly k templates will be constructed per image. When k is greater than the number of successful detections, additional Empty Template will be constructed.
Rotate images using roc_rotate on face detection failure. See the ImageRotation template metadata field in order to correct original image orientation for face bounding box display purposes.
Print to stderr the number of faces detected in each image.
0 - Don't log face count
1 - Log images with no faces detected
2 - Log face count for all images.
$ roc-represent -h
face.jpg to a gallery face.t. $ roc-represent face.jpg face.t
images/* to a gallery faces.t. $ roc-represent -r images/ faces.t
faces.t. $ ls -1 images/* | roc-represent faces.t
$ ls -1 images/* | roc-represent > faces.t
5 faces in faces.jpg to a gallery faces.t. $ roc-represent -k 5 faces.jpg faces.t
right_eye_x,right_eye_y,left_eye_x,left_eye_y,chin_x,chin_y . See roc_landmarks_to_detection. $ roc-represent --landmarks 505,227,581,236,535,373 face.jpg face.t
$ roc-represent -b /path/to/file.csv faces.t
{00000000-0000-0000-0000-000000000000}.{00000000-0000-0000-0000-000000000000}.RotatedFace_X,RotatedFace_Y,RotatedFace_Width,RotatedFace_Height.Face_X,Face_Y,Face_Width,Face_Height. .csv file example: 
1.8.15