Application API

APIs are specific to certain applications. User can add more

RGB Face Recognition

Start SFID mode

int kdp_start_sfid_mode(int dev_idx, uint32_t* img_size, float thresh, uint16_t width, uint16_t height, uint32_t format);

• start the user sfid mode with specified threshold, image format
• dev_idx: connected device ID. A host can connect several devices
• img_size: the required image file size will be returned.
• thresh: the threshold used to match face recognition result. if 0, the default threshold is used.
• img_width: the width of input image
• img_height: the height of input image
• format: the input image format
• return 0 on succeed, error code on failure

Face Recognition with User ID

int kdp_verify_user_id_generic(int dev_idx, uint16_t* user_id, char* img_buf, int buf_len, uint16_t* mask, char* res);

• extract the face feature from input image, and compare it with DB,
• return the matched user ID, and face analysis related result.
• dev_idx: connected device ID. A host can connect several devices
• user_id: matched user ID in DB will be returned.
• img_buf, buf_len: the image buffer and file size
• mask: indicate the requested and responsed flags
• bit 0 - FD result
• bit 1 - LM data
• bit 2 - FM feature map
• bit 3 - liveness
• res: contains all the related result
• FD/LM/FR/LV result size: please refer to model_res.h
• return 0 on succeed, error code on failure

Face Recognition with Feature Map

int kdp_extract_feature_generic(int dev_idx, char* img_buf, int buf_len, uint16_t* mask, char* res);

• extract face feature from inut image and save it in device,
• returns face detection, face recoginition result.
• dev_idx: connected device ID. A host can connect several devices
• img_buf, buf_len: the image buffer and file size
• mask: indicate the requested and responsed flags bit 0 - FD result bit 1 - LM data bit 2 - FM feature map bit 3 - liveness
• res: contains all the related result
• FD/LM/FR/LV result size: please refer to model_res.h
• return 0 on succeed, error code on failure

Get Result Size

uint32_t kdp_get_res_size(bool fd, bool lm, bool fr, bool liveness);

• @brief get result size
• @return result size

Get FDR Result Mask

uint16_t kdp_get_res_mask(bool fd, bool lm, bool fr, bool liveness);

• @brief get FDR result mask
• @return mask

Start Register Mode

int kdp_start_reg_user_mode(int dev_idx, uint16_t usr_id, uint16_t img_idx);

• start the user register mode
• dev_idx: connected device ID. A host can connect several devices
• user_id: the user id that will be registered.
• img_idx: the image idx that will be saved. (a user could have 5 images)
• return 0 on succeed, -1 else

Register New User

int kdp_register_user(int dev_idx, uint32_t user_id);

• register the face features to device DB
• dev_idx: connected device ID. A host can connect several devices
• user_id: the user id that be registered. must be same as the kdp_start_reg_user_mode.
• before calling this API, host must have called kdp_extract_feature
• at least once successfully, otherwise, no features could be saved.
• return 0 on succeed, error code on failure

Delete User

int kdp_remove_user(int dev_idx, uint32_t user_id);

• remove user from device DB
• It needs to be called after start lw3d mode or start verify mode
• dev_idx: connected device ID. A host can connect several devices
• user_id: the user to be removed. 0 for all users
• return 0 on succeed, error code on failure

Compare Two Feature Map

float kdp_fm_compare(float *user_fm_a, float *user_fm_b, size_t fm_size);

• @brief Calculate similarity of two feature points
• @param user_fm_a buffer A of user feature map data
• @param user_fm_b buffer B of user feature map data
• @param fm_size size of user feature map data
• @return similarity score,smaller score are more similar
• errcode -1:parameter error
• @note user must ensure buffer A and B are the same size

Light Weight 3D Face Recognition

Start LW3D Mode

int kdp_start_lw3d_mode(int dev_idx, uint32_t* rgb_size, uint32_t* nir_size, float rgb_thresh, float nir_thresh, uint16_t rgb_width, uint16_t rgb_height, uint16_t nir_width, uint16_t nir_height, uint32_t rgb_fmt, uint32_t nir_fmt);

• start the light weight 3D mode with specified threshold, img format
• dev_idx: connected device ID. A host can connect several devices
• rgb_size: the required rgb image file size will be returned.
• nir_size: the required nir image file size will be returned.
• rgb_thresh: the threshold used to match rgb face recognition result. if 0, the default threshold is used.
• nir_thresh: the threshold used to match nir face recognition result. if 0, the default threshold is used.
• rgb_width: the width of rgb image
• rgb_height: the height of rgb image
• nir_width: the width of nir image
• nir_height: the height of nir image
• rgb_fmt: the format of rgb image
• nir_fmt: the format of nir image
• return 0 on succeed, error code on failure

LW3D Inference with User ID

int kdp_verify_lw3D_image_generic(int dev_idx, uint16_t* user_id, char* img_buf, int buf_len, char* img_buf_nir, int buf_len_nir, uint16_t* mask, char* res);

• extract the face feature from input rgb/nir image, and compare it with DB,
• return the matched user ID, and face analysis related result.
• dev_idx: connected device ID. A host can connect several devices
• user_id: matched user ID in DB will be returned.
• img_buf, buf_len: the rgb image buffer and file size

• img_buf_nir, buf_len_nir: the nir image buffer and file size

• mask: indicate the requested and responsed flags bit 0 - FD result bit 1 - FM feature map bit 2 - LM data bit 4 - NIR FD result bit 5 - NIR feature map bit 6 - NIR LM data bit 8 - Final liveness

• res: contains all the related result
• FD/LM/FR/LV result size: please refer to model_res.h
• return 0 on succeed, error code on failure

LW3D Inference with User Feature Map

int kdp_extract_lw3D_feature_generic(int dev_idx, char* img_buf, int buf_len, char* img_buf_nir, int buf_len_nir, uint16_t* mask, char* res);

• extract the face feature from input rgb/nir image, and save it in device,
• return the face analysis related result if required.
• dev_idx: connected device ID. A host can connect several devices
• img_buf, buf_len: the rgb image buffer and file size

• img_buf_nir, buf_len_nir: the nir image buffer and file size

• mask: indicate the requested and responsed flags bit 0 - FD result bit 1 - FM feature map bit 2 - LM data bit 4 - NIR FD result bit 5 - NIR feature map bit 6 - NIR LM data bit 8 - Final liveness

• res: contains all the related result
• FD/LM/FR/LV result size: please refer to model_res.h
• return 0 on succeed, error code on failure

LW3D Result Mask

uint16_t kdp_get_lw3D_res_mask(bool fd, bool lm, bool fr, bool nir_fd, bool nir_lm, bool nir_fr, bool liveness);

• @brief get lw3D result mask
• @return mask

LW3D Result Size

uint32_t kdp_get_lw3D_res_size(bool fd, bool lm, bool fr, bool nir_fd, bool nir_lm, bool nir_fr, bool liveness);

• @brief get lw3D result size
• @return result size