Main modules

Katna.video module

class Katna.video.Video(autoflip_build_path=None, autoflip_model_path=None)

Bases: object

Class for all video frames operations

Parameters:object (class:Object) – base class inheritance

compress_video(file_path, force_overwrite=False, crf_parameter=23, output_video_codec='libx264', out_dir_path='', out_file_name='')

Function to compress given input file

Parameters:

• file_path (str) – Input file path
• force_overwrite (bool, optional) – optional parameter if True then if there is already a file in output file location function will overwrite it, defaults to False
• crf_parameter (int, optional) – Constant Rate Factor Parameter for controlling amount of video compression to be applied, The range of the quantizer scale is 0-51: where 0 is lossless, 23 is default, and 51 is worst possible. It is recommend to keep this value between 20 to 30 A lower value is a higher quality, you can change default value by changing config.Video.video_compression_crf_parameter
• output_video_codec (str, optional) – Type of video codec to choose, Currently supported options are libx264 and libx265, libx264 is default option. libx264 is more widely supported on different operating systems and platforms, libx265 uses more advanced x265 codec and results in better compression and even less output video sizes with same or better quality. Right now libx265 is not as widely compatible on older versions of MacOS and Widows by default. If wider video compatibility is your goal you should use libx264., you can change default value by changing Katna.config.Video.video_compression_codec
• out_dir_path (str, optional) – output folder path where you want output video to be saved, defaults to “”
• out_file_name (str, optional) – output filename, if not mentioned it will be same as input filename, defaults to “”

Raises:

Exception – raises FileNotFoundError Exception if input video file not found, also exception is raised in case output video file path already exist and force_overwrite is not set to True.

Returns:

Status code Returns True if video compression was successfull else False

Return type:

bool

compress_videos_from_dir(dir_path, force_overwrite=False, crf_parameter=23, output_video_codec='libx264', out_dir_path='', out_file_name='')

Function to compress input video files in a folder

Parameters:

• dir_path (str) – Input folder path
• force_overwrite (bool, optional) – optional parameter if True then if there is already a file in output file location function will overwrite it, defaults to False
• crf_parameter (int, optional) – Constant Rate Factor Parameter for controlling amount of video compression to be applied, The range of the quantizer scale is 0-51: where 0 is lossless, 23 is default, and 51 is worst possible. It is recommend to keep this value between 20 to 30 A lower value is a higher quality, you can change default value by changing config.Video.video_compression_crf_parameter
• output_video_codec (str, optional) – Type of video codec to choose, Currently supported options are libx264 and libx265, libx264 is default option. libx264 is more widely supported on different operating systems and platforms, libx265 uses more advanced x265 codec and results in better compression and even less output video sizes with same or better quality. Right now libx265 is not as widely compatible on older versions of MacOS and Widows by default. If wider video compatibility is your goal you should use libx264., you can change default value by changing Katna.config.Video.video_compression_codec
• out_dir_path (str, optional) – output folder path where you want output video to be saved, defaults to “”

Raises:

Exception – raises FileNotFoundError Exception if input video file not found, also exception is raised in case output video file path already exist and force_overwrite is not set to True.

Returns:

Status code Returns True if video compression was successfull else False

Return type:

bool

extract_keyframes_from_videos_dir(no_of_frames, dir_path, writer)

Returns best key images/frames from the videos in the given directory. you need to mention number of keyframes as well as directory path containing videos. Function returns python dictionary with key as filepath each dictionary element contains list of python numpy image objects as keyframes.

Parameters:

• no_of_frames (int, required) – Number of key frames to be extracted
• dir_path (str, required) – Directory location with videos
• writer (Writer, required) – Writer class obj to process keyframes

Returns:

Dictionary with key as filepath and numpy.2darray Image objects

Return type:

dict

extract_video_keyframes(no_of_frames, file_path, writer)

Returns a list of best key images/frames from a single video.

Parameters:

• no_of_frames (int, required) – Number of key frames to be extracted
• file_path (str, required) – video file location
• writer (Writer, required) – Writer object to process keyframe data

Returns:

List of numpy.2darray Image objects

Return type:

list

extract_video_keyframes_big_video(no_of_frames, file_path)

Parameters:

• no_of_frames –
• file_path –

Returns:

Return type:

resize_video(file_path, abs_file_path_output, aspect_ratio)

Resize a single video file

Parameters:

• file_path (str) – file path of the video to be resized
• abs_file_path_output (str) – absolute path to output video file
• aspect_ratio ([type]) – aspect ratio of the final video

Raises:

Exception – [description]

resize_video_from_dir(dir_path, abs_dir_path_output, aspect_ratio)

Resize all videos inside the directory

Parameters:

• dir_path (str) – Directory path where videos are located
• abs_dir_path_output (str) – Absolute path to directory where output videos should to be dumped
• aspect_ratio ([type]) – desirable aspect ratio for the videos

Raises:

Exception – [description]

save_frame_to_disk(frame, file_path, file_name, file_ext)

saves an in-memory numpy image array on drive.

Parameters:

• frame (numpy.ndarray, required) – In-memory image. This would have been generated by extract_video_keyframes method
• file_name (str, required) – name of the image.
• file_path (str, required) – Folder location where files needs to be saved
• file_ext (str, required) – File extension indicating the file type for example - ‘.jpg’

Returns:

None

Katna.image module

class Katna.image.Image(disable_text=True)

Bases: object

Class for all image cropping operations

Parameters:object (class:Object) – base class inheritance

Constructor for image files

crop_image(file_path, crop_width, crop_height, num_of_crops, writer, filters=[], down_sample_factor=8)

smartly crops the imaged based on the specification - width and height

Parameters:

• file_path (str, required) – Input file path
• crop_width (int) – output crop width
• crop_height (int) – output crop heigh
• num_of_crops (int) – number of crops required
• writer (Writer, required) – writer object to process data
• filters (list (eg. ['text'])) – filters to be applied for cropping(checks if image contains english text and the crop rectangle doesn’t cut the text)
• down_sample_factor (int [default=8]) – number by which you want to reduce image height & width (use it if image is large or to fasten the process)

Returns:

crop list

Return type:

list of structure crop_rect

crop_image_from_cvimage(input_image, crop_width, crop_height, num_of_crops, filters=[], down_sample_factor=8)

smartly crops the imaged based on the specification - width and height

Parameters:

• input_image (numpy array, required) – Input image
• crop_width (int) – output crop width
• crop_height (int) – output crop heigh
• num_of_crops (int) – number of crops required
• filters (list (eg. ['text'])) – filters to be applied for cropping(only returns crops containing english text where the crop rectangle doesn’t cut the text)
• down_sample_factor (int [default=8]) – number by which you want to reduce image height & width (use it if image is large or to fasten the process)

Returns:

crop list

Return type:

list of structure crop_rect

crop_image_from_dir(dir_path, crop_width, crop_height, num_of_crops, writer, filters=[], down_sample_factor=8)

smartly crops all the images (inside a directory) based on the specification - width and height

Parameters:

• dir_path (str, required) – Input Directory path
• crop_width (int) – output crop width
• crop_height (int) – output crop height
• num_of_crops (int) – number of crops required
• writer (int) – number of crops required
• filters (list (eg. ['text'])) – filters to be applied for cropping(checks if image contains english text and the crop rectangle doesn’t cut the text)
• down_sample_factor (int [default=8]) – number by which you want to reduce image height & width (use it if image is large or to fasten the process)

Returns:

crop dict with key as filepath and crop list for the file

Return type:

dict

crop_image_with_aspect(file_path, crop_aspect_ratio, num_of_crops, writer, filters=[], down_sample_factor=8)

smartly crops the imaged based on the aspect ratio and returns number of specified crops for each crop spec found in the image with the specified aspect ratio

Parameters:

• file_path (str, required) – Input file path
• crop_aspect_ratio (str (eg. '4:3')) – output crop ratio
• num_of_crops (Writer, required) – number of crops required
• filters (list (eg. ['text'])) – filters to be applied for cropping(checks if image contains english text and the crop rectangle doesn’t cut the text)
• down_sample_factor (int [default=8]) – number by which you want to reduce image height & width (use it if image is large or to fasten the process)
• writer – writer to process the image

Returns:

crop list

Return type:

list of structure crop_rect

resize_image(file_path, target_width, target_height, down_sample_factor=8)

smartly resizes the image based on the specification - width and height

Parameters:

• file_path (str, required) – Input file path
• target_width (int) – output image width
• target_height (int) – output image height
• down_sample_factor (int [default=8]) – number by which you want to reduce image height & width (use it if image is large or to fasten the process)

Returns:

resized image

Return type:

cv_image

resize_image_from_dir(dir_path, target_width, target_height, down_sample_factor=8)

smartly resizes all the images (inside a directory) based on the specification - width and height

Parameters:

• dir_path (str, required) – Input Directory path
• target_width (int) – output width
• target_height (int) – output height
• down_sample_factor (int [default=8]) – number by which you want to reduce image height & width (use it if image is large or to fasten the process)

Returns:

dict with key as filepath and resized image as in opencv format as value

Return type:

dict

save_crop_to_disk(crop_rect, frame, file_path, file_name, file_ext, rescale=False)

saves an in-memory crop on drive.

Parameters:

• crop_rect (crop_rect, required) – In-memory crop_rect.
• frame (numpy.ndarray, required) – In-memory input image.
• file_name (str, required) – name of the image.
• file_path (str, required) – Folder location where files needs to be saved
• file_ext (str, required) – File extension indicating the file type for example - ‘.jpg’

Returns:

None

save_image_to_disk(image, file_path, file_name, file_ext)

saves an in-memory image obtained from image resize on drive.

Parameters:

• image (numpy.ndarray, required) – In-memory input image.
• file_name (str, required) – name of the image.
• file_path (str, required) – Folder location where files needs to be saved
• file_ext (str, required) – File extension indicating the file type for example - ‘.jpg’

Returns:

None

class Katna.image.UserFiltersEnum

Bases: object

Enum class for filters

text = 'TextDetector'

Config module

class Katna.config.CropScorer

Bases: object

detail_weight = 0.2

edge_radius = 0.4

edge_weight = -20

face_bias = 0.01

face_weight = 3.4

outside_importance = -0.5

rects_weight = 1

rule_of_thirds = True

saliency_bias = 0.2

saliency_weight = 1.3

class Katna.config.EdgeFeature

Bases: object

ksize = 3

max_val_threshold = 200

min_val_threshold = 100

class Katna.config.FaceFeature

Bases: object

cache_subdir = 'models'

confidence = 0.5

model_file = 'res10_300x300_ssd_iter_140000_fp16.caffemodel'

modelfile_download_link = 'https://raw.githubusercontent.com/opencv/opencv_3rdparty/dnn_samples_face_detector_20180205_fp16/res10_300x300_ssd_iter_140000_fp16.caffemodel'

prototxt_download_link = 'https://raw.githubusercontent.com/opencv/opencv/master/samples/dnn/face_detector/deploy.prototxt'

prototxt_file = 'deploy.prototxt'

class Katna.config.FrameExtractor

Bases: object

USE_LOCAL_MAXIMA = True

len_window = 20

max_frames_in_chunk = 500

window_type = 'hanning'

class Katna.config.Image

Bases: object

DEBUG = False

crop_height_reduction_factor_in_each_iteration = 0.05

down_sample_factor = 8

min_image_to_crop_factor = 4

class Katna.config.ImageSelector

Bases: object

brightness_step = 2.0

entropy_step = 0.5

max_brightness_value = 90.0

max_entropy_value = 10.0

min_brightness_value = 10.0

min_entropy_value = 1.0

class Katna.config.MediaPipe

Bases: object

class AutoFlip

Bases: object

BLUR_AREA_OPACITY = 0.6

BLUR_AREA_OPACITY_KEYNAME = 'BLUR_AREA_OPACITY'

BUILD_CMD = 'run_autoflip'

CONFIG_FILE_PBTXT = '/home/docs/checkouts/readthedocs.org/user_builds/katna/checkouts/latest/Katna/mediapipe_autoflip.pbtxt'

DEFAULT_BLUR_AREA_OPACITY = 0.6

DEFAULT_FEATURE_SIGNAL_VALUE = 'false'

DEFAULT_MOTION_STABALIZATION_THRESHOLD = 0.5

ENFORCE_FEATURES = {'CAR': False, 'FACE_ALL_LANDMARKS': False, 'FACE_CORE_LANDMARKS': False, 'FACE_FULL': False, 'HUMAN': False, 'OBJECT': False, 'PET': False}

ENFORCE_FEATURES_KEYNAME = 'ENFORCE_FEATURES'

MODELS_FOLDER_LOCATION = '/home/docs/checkouts/readthedocs.org/user_builds/katna/checkouts/latest/docs/source/mediapipe/models'

RERUN_LIMIT = 2

STABALIZATION_THRESHOLD = 0.5

STABALIZATION_THRESHOLD_KEYNAME = 'STABALIZATION_THRESHOLD'

TMP_PBTXT_FOLDER_NAME = 'temp_pbtxt'

TMP_PBTXT_FOLDER_PATH = '/home/docs/checkouts/readthedocs.org/user_builds/katna/checkouts/latest/docs/source/temp_pbtxt'

classmethod get_conf()

Gets the current config

Returns:dictionary containing the current config Return type:dict

classmethod get_pbtxt_mapping()

classmethod set_conf(config)

Sets the config passed

Parameters:config (dict) – The configuration to set.

class Katna.config.TextDetector

Bases: object

cache_subdir = 'models'

frozen_weights = 'frozen_east_text_detection.pb'

layerNames = ['feature_fusion/Conv_7/Sigmoid', 'feature_fusion/concat_3']

merge_threshold = 1

min_confidence = 0.9

model_download_link = 'https://github.com/oyyd/frozen_east_text_detection.pb/raw/master/frozen_east_text_detection.pb'

class Katna.config.Video

Bases: object

DEBUG = False

assumed_no_of_frames_per_candidate_frame = 5

compression_output_file_extension = 'mp4'

memory_consumption_threshold = 0.8

min_video_duration = 5.0

video_compression_codec = 'libx264'

video_compression_crf_parameter = 23

video_extensions = ['.str', '.aa', '.aac', '.ac3', '.acm', '.adf', '.adp', '.dtk', '.ads', '.ss2', '.adx', '.aea', '.afc', '.aix', '.al', '.ape', '.apl', '.mac', '.aptx', '.aptxhd', '.aqt', '.ast', '.avi', '.avr', '.bfstm', '.bcstm', '.bit', '.bmv', '.brstm', '.cdg', '.cdxl', '.xl', '.c2', '.302', '.daud', '.str', '.dss', '.dts', '.dtshd', '.dv', '.dif', '.cdata', '.eac3', '.paf', '.fap', '.flm', '.flac', '.flv', '.fsb', '.g722', '.722', '.tco', '.rco', '.g723_1', '.g729', '.genh', '.gsm', '.h261', '.h26l', '.h264', '.264', '.avc', '.hevc', '.h265', '.265', '.idf', '.cgi', '.sf', '.ircam', '.ivr', '.flv', '.lvf', '.m4v', '.mkv', '.mk3d', '.mka', '.mks', '.mjpg', '.mjpeg', '.mpo', '.j2k', '.mlp', '.mov', '.mp4', '.m4a', '.3gp', '.3g2', '.mj2', '.mp2', '.mp3', '.m2a', '.mpa', '.mpc', '.mjpg', '.txt', '.mpl2', '.sub', '.msf', '.mtaf', '.ul', '.musx', '.mvi', '.mxg', '.v', '.nist', '.sph', '.nsp', '.nut', '.ogg', '.oma', '.omg', '.aa3', '.pjs', '.pvf', '.yuv', '.cif', '.qcif', '.rgb', '.rt', '.rsd', '.rsd', '.rso', '.sw', '.sb', '.smi', '.sami', '.sbc', '.msbc', '.sbg', '.scc', '.sdr2', '.sds', '.sdx', '.shn', '.vb', '.son', '.sln', '.mjpg', '.stl', '.sub', '.sub', '.sup', '.svag', '.tak', '.thd', '.tta', '.ans', '.art', '.asc', '.diz', '.ice', '.nfo', '.txt', '.vt', '.ty', '.ty+', '.uw', '.ub', '.v210', '.yuv10', '.vag', '.vc1', '.viv', '.idx', '.vpk', '.txt', '.vqf', '.vql', '.vqe', '.vtt', '.wsd', '.xmv', '.xvag', '.yop', '.y4m']

video_split_threshold_in_minutes = 20

Helper modules

Katna.decorators module

class Katna.decorators.DebugDecorators

Bases: object

File validation decorator

Arguments:

object {[type]} – [description]

Raises:

None:

Returns:

[] – [Decorated function]

classmethod add_optional_debug_images_for_image_module(decorated)

Add optional debug images in image_module class if DEBUG option is True in config

class Katna.decorators.FileDecorators

Bases: object

File validation decorator

Raises:FileNotFoundError – File or path is incorrect

classmethod validate_dir_path(decorated)

Validate if the input path is a valid dir or location

Parameters:decorated (function, required) – decorated function Returns:function if the path is valid Return type:function object

classmethod validate_file_path(decorated)

Validate if the input path is a valid file or location

Parameters:decorated (function, required) – decorated function Returns:function if the path is valid Return type:function object

class Katna.decorators.VideoDecorators

Bases: object

File validation decorator

Arguments:

object {[type]} – [description]

Raises:

FileNotFoundError: [Video File is missing]

Returns:

[boolean] – [if the file exists and is valid]

classmethod validate_video(decorated)

Validate if the input video is a valid file

Katna.decorators.exception(logger)

A decorator that wraps the passed in function and logs exceptions should one occur

param logger: The logging object type logger: logger

Internal modules for Video

Katna.frame_extractor module

class Katna.frame_extractor.FrameExtractor

Bases: object

Class for extraction of key frames from video : based on sum of absolute differences in LUV colorspace from given video

extract_candidate_frames(videopath)

Pubic function for this module , Given and input video path This functions Returns one list of all candidate key-frames

Parameters:

• object (class:Object) – base class inheritance
• videopath (str) – inputvideo path

Returns:

opencv.Image.Image objects

Return type:

list

Katna.image_selector module

class Katna.image_selector.ImageSelector(n_processes=1)

Bases: object

Class for selection of best top N images from input list of images, Currently following selection method are implemented: brightness filtering, contrast/entropy filtering, clustering of frames and variance of laplacian for non blurred images selection

Parameters:object (class:Object) – base class inheritance

select_best_frames(input_key_frames, number_of_frames)

[summary] Public function for Image selector class: takes list of key-frames images and number of required frames as input, returns list of filtered keyframes

Parameters:

• object (class:Object) – base class inheritance
• input_key_frames (python list opencv images) – list of input keyframes in list of opencv image format
• number_of_frames – Required number of images

Type:

int

Returns:

Returns list of filtered image files

Return type:

python list of images

Katna.video_compressor module

class Katna.video_compressor.VideoCompressor

Bases: object

Class for performing video compression task: based on ffmpeg library for doing video compression

compress_video(file_path, force_overwrite, crf_parameter, output_video_codec, out_dir_path, out_file_name)

Function to compress given input file

Parameters:

• file_path (str) – Input file path
• force_overwrite (bool, optional) – optional parameter if True then if there is already a file in output file location function will overwrite it, defaults to False
• crf_parameter (int, optional) – Constant Rate Factor Parameter for controlling amount of video compression to be applied, The range of the quantizer scale is 0-51: where 0 is lossless, 23 is default, and 51 is worst possible. It is recommend to keep this value between 20 to 30 A lower value is a higher quality, you can change default value by changing config.Video.video_compression_crf_parameter
• output_video_codec (str, optional) – Type of video codec to choose, Currently supported options are libx264 and libx265, libx264 is default option. libx264 is more widely supported on different operating systems and platforms, libx265 uses more advanced x265 codec and results in better compression and even less output video sizes with same or better quality. Right now libx265 is not as widely compatible on older versions of MacOS and Widows by default. If wider video compatibility is your goal you should use libx264., you can change default value by changing Katna.config.Video.video_compression_codec
• out_dir_path (str, optional) – output folder path where you want output video to be saved, defaults to “”
• out_file_name (str, optional) – output filename, if not mentioned it will be same as input filename, defaults to “”

Raises:

Exception – raises FileNotFoundError Exception if input video file not found, also exception is raised in case output video file path already exist and force_overwrite is not set to True.

Returns:

Status code Returns True if video compression was successfull else False

Return type:

bool

Internal modules for Image

Katna.crop_rect module

class Katna.crop_rect.CropRect(x, y, w, h)

Bases: object

Data structure class for storing image crop rectangles

Parameters:object (class:Object) – base class inheritance

get_image_crop(input_image)

public functions which for given input image and current crop rectangle object returns image cropped to crop rectangle specifications.

Parameters:

• object (class:Object) – base class inheritance
• image (Opencv Numpy Image) – input image

Returns:

cropped image according to given spec

Return type:

Opencv Numpy Image

Katna.crop_extractor module

class Katna.crop_extractor.CropExtractor

Bases: object

Class for extraction and scoring of all possible crops from input image for input crop size. Currently features being used for scoring a given crop retangle are: Edge, saliency and face detection. Additionally crop scoring also takes following metrics into account: Distance of pixel in crop rectangle from crop boundary and rule of third.

extract_candidate_crops(inputImage, crop_width, crop_height, feature_list)

Public Function for crop_extractor module, Given input image and desired crop width and height: function returns list of all candidate crop rectangles scored taking into account input Image Feature list

Parameters:

• object (class:Object) – base class inheritance
• inputImage (opencv.Image) – input Image
• crop_width (Int) – input crop width
• crop_height (Int) – input crop height
• feature_list (List of OpenCV numpy image type) – list of input feature maps to be used for scoring a crop rectangle

Returns:

List of extracted crop rectangle objects

Return type:

list of crop_rect

Katna.crop_selector module

class Katna.crop_selector.CropSelector

Bases: object

Class for selection of top n crop rectangles from input list of crop rectangles ,It also implements filtering functionality, currently following filtering method are implemented: Text Detector filter

Parameters:object (class:Object) – base class inheritance

select_candidate_crops(input_image, num_of_crops, input_crop_list, defined_filters, filters=[])

Public Function for CropSelector class: takes list of crop rectangles and number of required crops as input and returns list of filtered crop retangles as output. Also takes list of filters to be used for filtering out unwanted crop rectangles as optional input.

Parameters:

• object (class:Object) – base class inheritance
• input_image (Opencv Numpy Image) – input image
• input_crop_list (python list crop_rect data structure) – list of input crop in list of crop_rect format
• number_of_crop – Required number of crop

Type:

int

Returns:

Returns list of filtered crop_rect

Return type:

python list of crop_rect data structure

Image Filters for Smart Cropping

Katna.image_filters.filter

class Katna.image_filters.filter.Filter(weight)

Bases: object

Base Class for Image filters

get_weight()

gets weight of particular filter

Returns:weight of the filter Return type:float

Katna.image_filters.text_detector

class Katna.image_filters.text_detector.TextDetector(weight=1.0)

Bases: Katna.image_filters.filter.Filter

TextDetector Class: Class for implementation of text detector filter, inherit from Filter class

Constructor for this class does following tasks, if not already downloaded , it first downloads text detector dnn weights file from public URL ands save it at USER_HOME/.katna directory, or /tmp/.katna directory. After this initializer code initializes internal parameter: min_confidence (for text detection)

download_data()

Public function for downloading the network weight from the URL link, to be used for text detection functionality. Troubleshooting tip: If you get FileNotFound error during text detector initialization, initialize the text detector and call this function directly to download the model file from public URL link.

get_filter_result(crop)

Main public function of TextDetector filter class, this filter Returns false if crop contains no text, additionally checks for overlap between input crop rectangle and the detected text bounding box, returns True if No overlap (Filter will not discard input crop) otherwise returns False (signal for discarding input crop).

Parameters:crop (crop_rect) – input crop rectangle to test Returns:True if No overlap (Filter will not discard input crop) otherwise returns False Return type:bool

set_image(image)

Public set_image function, This will detect all text boxes in input image and will saves them as internal list of text_rect to be used in get_filter_result

Parameters:image (numpy array(opencv)) – input image from which needs to be cropped

Image Features for Smart Cropping

Katna.image_features.feature

class Katna.image_features.feature.Feature(weight)

Bases: object

Base Class: Base class for all Image features: To be used for Katna crop rectangle scoring, all image features inherit from this base class

get_weight()

gets weight of particular feature

Returns:weight of the feature Return type:float

Katna.image_features.edge_feature

class Katna.image_features.edge_feature.EdgeFeature(weight=0.0)

Bases: Katna.image_features.feature.Feature

Class for getting edge detector Feature, Constructor Parameters:- feature weight

Internal Parameters:-

min_val_threshold : min edge threshold

max_val_threshold : max edge threshold

ksize: size of Sobel kernel used for finding image gradients

get_feature_map(image)

Public function for getting Feature map image from edges detection

Parameters:image (numpy array) – input image Returns:single channel opencv numpy image with feature map from edge detection Return type:numpy array

Katna.image_features.saliency_feature

class Katna.image_features.saliency_feature.SaliencyFeature(weight=0.0)

Bases: Katna.image_features.feature.Feature

Class for calculating saliency feature map from Input image

get_feature_map(image)

Public function for getting Feature map image from Image saliency detection

Parameters:image (numpy array) – input image Returns:single channel opencv numpy image with feature map from saliency detection Return type:numpy array

Katna.image_features.face_feature

class Katna.image_features.face_feature.FaceFeature(weight=1.0)

Bases: Katna.image_features.feature.Feature

Class for calculating Face detection feature map from input image

Internal Parameters

model_file : Path for downloading model for face detection

prototxt_file : Path for downloading model description prototxt file

confidence : Face detection confidence threshold value

Constructor for this class does following tasks, if not already downloaded , it first downloads face detector model file and prototxt file from public URL ands save it at USER_HOME/.katna directory, or /tmp/.katna directory. After this initializer code initializes internal parameter: min_confidence (for face detection)

download_model()

Public function for downloading the network weight from the URL link, to be used for face detection functionality. Troubleshooting tip: If you get FileNotFound error during face detector initialization, initialize the face detector and call this function directly to download the model file from public URL link.

download_proto()

Public function for downloading the network weight from the URL link, to be used for face detection functionality. Troubleshooting tip: If you get FileNotFound error during face detector initialization, initialize the face detector and call this function directly to download the model file from public URL link.

get_feature_map(image)

Public function for getting Feature map image from Face detection in input Image

Parameters:image (numpy array) – input image Returns:single channel opencv numpy image with feature map from Face detection Return type:numpy array