pyapi

script

Script

class lost.pyapi.script.Script(pe_id=None)[source]

Superclass for a user defined Script.

Custom scripts need to inherit from Script and implement the main method.

pe_id

Pipe element id. Assign the pe id of a pipline script in order to emulate this script in a jupyter notebook for example.

Type:int
break_loop()[source]

Break next loop in pipeline.

create_label_tree(name, external_id=None)[source]

Create a new LabelTree

Parameters:
  • name (str) – Name of the tree / name of the root leaf.
  • external_id (str) – An external id for the root leaf.
Returns:

The created LabelTree.

Return type:

lost.logic.label.LabelTree

get_abs_path(path)[source]

Get absolute path in current file system.

Parameters:path (str) – A relative path.
Returns:Absolute path
Return type:str
get_alien_element(pe_id)[source]

Get an pipeline element by id from somewhere in the LOST system.

It is an alien element since it is most likely not part of the pipeline instance this script belongs to.

Parameters:pe_id (int) – PipeElementID of the alien element.
Returns:
get_arg(arg_name)[source]

Get argument value by name for this script.

Parameters:arg_name (str) – Name of the argument.
Returns:Value of the given argument.
get_label_tree(name)[source]

Get a LabelTree by name.

Parameters:name (str) – Name of the desired LabelTree.
Retruns:
lost.logic.label.LabelTree or None:
If a label tree with the given name exists it will be returned. Otherwise None will be returned
get_path(file_name, context='instance', ptype='abs')[source]

Get path for the filename in a specific context in filesystem.

Parameters:
  • file_name (str) – Name or relative path for a file.
  • context (str) – Options: instance, pipe, static
  • ptype (str) – Type of this path. Can be relative or absolute Options: abs, rel
Returns:

Path to the file in the specified context.

Return type:

str

get_rel_path(path)[source]

Get relativ path for current project

Parameters:path (str) – A absolute path
Returns:Relative path
Return type:str
inp

lost.pyapi.inout.Input

instance_context

Get the path to store files that are only valid for this instance.

Type:str
iteration

Get the current iteration.

Number of times this script has been executed.

Type:int
logger

A standard python logger for this script.

It will log to the pipline log file.

Type:logging.Logger
loop_is_broken()[source]

Check if the current loop is broken

outp

lost.pyapi.inout.ScriptOutput

pipe_context

Root path to store files that should be visible for all elements in the pipeline.

Type:str
pipe_info

An object with pipeline informations

Type:lost.pyapi.pipeline.PipeInfo
progress

Get current progress that is displayed in the progress bar of this script.

Current progress in percent 0…100

Type:float
reject_execution()[source]

Reject execution of this script and set it to PENDING again.

Note

This method is useful if you want to execute this script only when some condition based on previous pipeline elements is meet.

report_err(msg)[source]

Report an error for this user script to portal

Parameters:msg – The error message that should be reported.

Note

You can call this method multiple times if you like. All messages will be concatenated an sent to the portal.

static_context

Get the static path.

Files that are stored at this path can be accessed by all instances of a script.

Type:str
update_progress(value)[source]

Update the progress for this script.

Parameters:value (float) – Progress in percent 0…100

inout

ScriptOutput

class lost.pyapi.inout.ScriptOutput(script)[source]

Special Output class since lost.pyapi.script.Script objects may manipulate and request annotations.

add_annos(img_path, img_labels=None, img_sim_class=None, annos=[], anno_types=[], anno_labels=[], anno_sim_classes=[], frame_n=None, video_path=None)[source]

Add annos in list style to an image.

Parameters:
  • img_path (str) – Path to the image where annotations are added for.
  • img_labels (list of int) – Labels that will be assigned to the image. Each label in the list is represented by a label_leaf_id.
  • img_sim_class (int) – A culster id that will be used to cluster this image in the MIA annotation tool.
  • annos (list of list) – A list of POINTs: [x,y] BBOXes: [x,y,w,h] LINEs or POLYGONs: [[x,y], [x,y], …]
  • anno_types (list of str) – Can be ‘point’, ‘bbox’, ‘line’, ‘polygon’
  • anno_labels (list of list of int) – Labels for the twod annos. Each label in the list is represented by a label_leaf_id. (see also LabelLeaf).
  • anno_sim_classes (list of ints) – List of arbitrary cluster ids that are used to cluster annotations in the MIA annotation tool.
  • frame_n (int) – If img_path belongs to a video frame_n indicates the framenumber.
  • video_path (str) – If img_path belongs to a video this is the path to this video.

Example

Add annotations to an:

>>> self.outp.add_annos('path/to/img.jpg',
...     annos = [
...         [0.1, 0.1, 0.2, 0.2],
...         [0.1, 0.2],
...         [[0.1, 0.3], [0.2, 0.3], [0.15, 0.1]]
...     ],
...     anno_types=['bbox', 'point', 'polygon'],
...     anno_labels=[
...         [1],
...         [1],
...         [4]
...     ],
...     anno_sim_classes=[10, 10, 15]
... )

Note

In contrast to request_annos this method will broadcast the added annotations to all connected pipeline elements.

add_data_export(file_path)[source]

Serve a file for download inside the web gui via a DataExport element.

Parameters:file_path (str) – Path to the file that should be provided for download.
add_visual_output(img_path=None, html=None)[source]

Display an image and html in the web gui via a VisualOutput element.

Parameters:
  • img_path (str) – Path in the lost filesystem to the image to display.
  • html (str) – HTML text to display.
anno_tasks

list of lost.pyapi.pipe_elements.AnnoTask objects

bbox_annos

Iterate over all bbox annotation.

Returns:Iterator of lost.db.model.TwoDAnno.
data_exports

list of lost.pyapi.pipe_elements.VisualOutput objects.

datasources

list of lost.pyapi.pipe_elements.Datasource objects

img_annos

Iterate over all lost.db.model.ImageAnno objects in this Resultset.

Returns:Iterator of lost.db.model.ImageAnno objects.
line_annos

Iterate over all line annotations.

Returns:Iterator of lost.db.model.TwoDAnno objects.
mia_tasks

list of lost.pyapi.pipe_elements.MIATask objects

point_annos

Iterate over all point annotations.

Returns:Iterator of lost.db.model.TwoDAnno.
polygon_annos

Iterate over all polygon annotations.

Returns:Iterator of lost.db.model.TwoDAnno objects.
raw_files

list of lost.pyapi.pipe_elements.RawFile objects

request_annos(img_path, img_labels=None, img_sim_class=None, annos=[], anno_types=[], anno_labels=[], anno_sim_classes=[], frame_n=None, video_path=None)[source]

Request annotations for a subsequent annotaiton task.

Parameters:
  • img_path (str) – Path to the image where annotations are added for.
  • img_label (list of int) – Labels that will be assigned to the image. The labels should be represented by a label_leaf_id. An image may have multiple labels.
  • img_sim_class (int) – A culster id that will be used to cluster this image in the MIA annotation tool.
  • annos (list of list) – A list of POINTs: [x,y] BBOXes: [x,y,w,h] LINEs or POLYGONs: [[x,y], [x,y], …]
  • anno_types (list of str) – Can be ‘point’, ‘bbox’, ‘line’, ‘polygon’
  • anno_labels (list of int) – Labels for the twod annos. Each label in the list is represented by a label_leaf_id. (see also LabelLeaf).
  • anno_sim_classes (list of ints) – List of arbitrary cluster ids that are used to cluster annotations in the MIA annotation tool.
  • frame_n (int) – If img_path belongs to a video frame_n indicates the framenumber.
  • video_path (str) – If img_path belongs to a video this is the path to this video.

Example

Request human annotations for an image with annotation proposals:

 >>> self.outp.add_annos('path/to/img.jpg',
...     annos = [
...         [0.1, 0.1, 0.2, 0.2],
...         [0.1, 0.2],
...         [[0.1, 0.3], [0.2, 0.3], [0.15, 0.1]]
...     ],
...     anno_types=['bbox', 'point', 'polygon'],
...     anno_labels=[
...         [1],
...         [1],
...         [4]
...     ],
...     anno_sim_classes=[10, 10, 15]
... )

Reqest human annotations for an image without porposals:

>>> self.outp.request_annos('path/to/img.jpg')
request_bbox_annos(img_path, boxes=[], labels=[], frame_n=None, video_path=None, sim_classes=[])[source]

Request BBox annotations for a subsequent annotaiton task.

Parameters:
  • img_path (str) – Path of the image.
  • boxes (list) – A list of boxes [[x,y,w,h],..].
  • labels (list) – A list of labels for each box.
  • frame_n (int) – If img_path belongs to a video frame_n indicates the framenumber.
  • video_path (str) – If img_path belongs to a video this is the path to this video.
  • sim_classes (list) – [sim_class1, sim_class2,…] A list of similarity classes that is used to cluster BBoxes when using MIA for annotation.

Note

There are three cases when you request a bbox annotation.

Case1: Annotate empty image
You just want to get bounding boxes drawn by a human annotator for an image. -> Only set the img_path argument.
Case2: Annotate image with a preset of boxes
You want to get verified predicted bounding boxes by a human annotator and you have not predicted a label for the boxes. -> Set the img_path argument and boxes.
Case3: Annotate image with a preset of boxes and labels
You want to get predicted bounding boxes and the related predicted labels to be verified by a human annotator. -> Set the img_path and the boxes argument. For boxes you need to assign a list of box and a list of label_ids for labels. An annotation may have multiple labels. E.g. boxes =[[0.1,0.1,0.2,0.3],…], labels =[[1,5],[5],…]

Example

How to use this method in a Script:

>>> self.request_bbox_annos('path/to/img.png',
...     boxes=[[0.1,0.1,0.2,0.3],[0.2,0.2,0.4,0.4]],
...     labels=[[0],[1]]
... )
request_image_anno(img_path, sim_class=None, labels=None, frame_n=None, video_path=None)[source]

Request a class label annotation for an image.

Parameters:
  • img_path (str) – Path to the image that should be annotated.
  • sim_class (int) – A similarity class for this image. This similarity measure will be used to cluster images for MultiObjectAnnoation -> Images with the same sim_class will be presented to the annotator in one step.
  • labels (list of int) – Labels that will be assigned to the image. Each label should represent a label_leaf_id.
  • frame_n (int) – If img_path belongs to a video frame_n indicates the framenumber.
  • video_path (str) – If img_path belongs to a video this is the path to this video.

Example

Request image annotation::
>>> self.request_image_anno('path/to/image', sim_class=2)
sia_tasks

list of lost.pyapi.pipe_elements.SIATask objects

to_df()

Get a pandas DataFrame of all annotations related to this object.

Returns:
Column names are:
’img.idx’, ‘img.anno_task_id’, ‘img.timestamp’, ‘img.timestamp_lock’, ‘img.state’, ‘img.sim_class’, ‘img.frame_n’, ‘img.video_path’, ‘img.img_path’, ‘img.result_id’, ‘img.iteration’, ‘img.group_id’, ‘img.anno_time’, ‘img.lbl.idx’, ‘img.lbl.name’, ‘img.lbl.external_id’, ‘img.annotator’, ‘anno.idx’, ‘anno.anno_task_id’, ‘anno.timestamp’, ‘anno.timestamp_lock’, ‘anno.state’, ‘anno.track_n’, ‘anno.dtype’, ‘anno.sim_class’, ‘anno.iteration’, ‘anno.group_id’, ‘anno.img_anno_id’, ‘anno.annotator’, ‘anno.confidence’, ‘anno.anno_time’, ‘anno.lbl.idx’, ‘anno.lbl.name’, ‘anno.lbl.external_id’, ‘anno.data’
Return type:pandas.DataFrame
to_vec(columns='all')

Get a vector of all Annotations related to this object.

Parameters:columns (str or list of str) – ‘all’ OR ‘img.idx’, ‘img.anno_task_id’, ‘img.timestamp’, ‘img.timestamp_lock’, ‘img.state’, ‘img.sim_class’, ‘img.frame_n’, ‘img.video_path’, ‘img.img_path’, ‘img.result_id’, ‘img.iteration’, ‘img.group_id’, ‘img.anno_time’, ‘img.lbl.idx’, ‘img.lbl.name’, ‘img.lbl.external_id’, ‘img.annotator’, ‘anno.idx’, ‘anno.anno_task_id’, ‘anno.timestamp’, ‘anno.timestamp_lock’, ‘anno.state’, ‘anno.track_n’, ‘anno.dtype’, ‘anno.sim_class’, ‘anno.iteration’, ‘anno.group_id’, ‘anno.img_anno_id’, ‘anno.annotator’, ‘anno.confidence’, ‘anno.anno_time’, ‘anno.lbl.idx’, ‘anno.lbl.name’, ‘anno.lbl.external_id’, ‘anno.data’
Retruns:
list OR list of lists: Desired columns

Example

Return just a list of 2d anno labels:

>>> img_anno.to_vec('anno.lbl.name')
['Aeroplane', 'Bicycle', 'Bottle', 'Horse']

Return a list of lists:

>>> self.inp.get_anno_vec.(['img.img_path', 'anno.lbl.name',
...     'anno.data', 'anno.dtype'])
[
    ['path/to/img1.jpg', 'Aeroplane', [0.1, 0.1, 0.2, 0.2], 'bbox'],
    ['path/to/img1.jpg', 'Bicycle', [0.1, 0.1], 'point'],
    ['path/to/img2.jpg', 'Bottle', [[0.1, 0.1], [0.2, 0.2]], 'line'],
    ['path/to/img3.jpg', 'Horse', [0.2, 0.15, 0.3, 0.18], 'bbox']
]
twod_annos

Iterate over 2D-annotations.

Returns:of lost.db.model.TwoDAnno objects.
Return type:Iterator
visual_outputs

list of lost.pyapi.pipe_elements.VisualOutput objects.

Output

class lost.pyapi.inout.Output(element)[source]
anno_tasks

list of lost.pyapi.pipe_elements.AnnoTask objects

bbox_annos

Iterate over all bbox annotation.

Returns:Iterator of lost.db.model.TwoDAnno.
data_exports

list of lost.pyapi.pipe_elements.VisualOutput objects.

datasources

list of lost.pyapi.pipe_elements.Datasource objects

img_annos

Iterate over all lost.db.model.ImageAnno objects in this Resultset.

Returns:Iterator of lost.db.model.ImageAnno objects.
line_annos

Iterate over all line annotations.

Returns:Iterator of lost.db.model.TwoDAnno objects.
mia_tasks

list of lost.pyapi.pipe_elements.MIATask objects

point_annos

Iterate over all point annotations.

Returns:Iterator of lost.db.model.TwoDAnno.
polygon_annos

Iterate over all polygon annotations.

Returns:Iterator of lost.db.model.TwoDAnno objects.
raw_files

list of lost.pyapi.pipe_elements.RawFile objects

sia_tasks

list of lost.pyapi.pipe_elements.SIATask objects

to_df()

Get a pandas DataFrame of all annotations related to this object.

Returns:
Column names are:
’img.idx’, ‘img.anno_task_id’, ‘img.timestamp’, ‘img.timestamp_lock’, ‘img.state’, ‘img.sim_class’, ‘img.frame_n’, ‘img.video_path’, ‘img.img_path’, ‘img.result_id’, ‘img.iteration’, ‘img.group_id’, ‘img.anno_time’, ‘img.lbl.idx’, ‘img.lbl.name’, ‘img.lbl.external_id’, ‘img.annotator’, ‘anno.idx’, ‘anno.anno_task_id’, ‘anno.timestamp’, ‘anno.timestamp_lock’, ‘anno.state’, ‘anno.track_n’, ‘anno.dtype’, ‘anno.sim_class’, ‘anno.iteration’, ‘anno.group_id’, ‘anno.img_anno_id’, ‘anno.annotator’, ‘anno.confidence’, ‘anno.anno_time’, ‘anno.lbl.idx’, ‘anno.lbl.name’, ‘anno.lbl.external_id’, ‘anno.data’
Return type:pandas.DataFrame
to_vec(columns='all')

Get a vector of all Annotations related to this object.

Parameters:columns (str or list of str) – ‘all’ OR ‘img.idx’, ‘img.anno_task_id’, ‘img.timestamp’, ‘img.timestamp_lock’, ‘img.state’, ‘img.sim_class’, ‘img.frame_n’, ‘img.video_path’, ‘img.img_path’, ‘img.result_id’, ‘img.iteration’, ‘img.group_id’, ‘img.anno_time’, ‘img.lbl.idx’, ‘img.lbl.name’, ‘img.lbl.external_id’, ‘img.annotator’, ‘anno.idx’, ‘anno.anno_task_id’, ‘anno.timestamp’, ‘anno.timestamp_lock’, ‘anno.state’, ‘anno.track_n’, ‘anno.dtype’, ‘anno.sim_class’, ‘anno.iteration’, ‘anno.group_id’, ‘anno.img_anno_id’, ‘anno.annotator’, ‘anno.confidence’, ‘anno.anno_time’, ‘anno.lbl.idx’, ‘anno.lbl.name’, ‘anno.lbl.external_id’, ‘anno.data’
Retruns:
list OR list of lists: Desired columns

Example

Return just a list of 2d anno labels:

>>> img_anno.to_vec('anno.lbl.name')
['Aeroplane', 'Bicycle', 'Bottle', 'Horse']

Return a list of lists:

>>> self.inp.get_anno_vec.(['img.img_path', 'anno.lbl.name',
...     'anno.data', 'anno.dtype'])
[
    ['path/to/img1.jpg', 'Aeroplane', [0.1, 0.1, 0.2, 0.2], 'bbox'],
    ['path/to/img1.jpg', 'Bicycle', [0.1, 0.1], 'point'],
    ['path/to/img2.jpg', 'Bottle', [[0.1, 0.1], [0.2, 0.2]], 'line'],
    ['path/to/img3.jpg', 'Horse', [0.2, 0.15, 0.3, 0.18], 'bbox']
]
twod_annos

Iterate over 2D-annotations.

Returns:of lost.db.model.TwoDAnno objects.
Return type:Iterator
visual_outputs

list of lost.pyapi.pipe_elements.VisualOutput objects.

Input

class lost.pyapi.inout.Input(element)[source]

Class that represants an input of a pipeline element.

Parameters:element (object) – Related lost.db.model.PipeElement object.
anno_tasks

list of lost.pyapi.pipe_elements.AnnoTask objects

bbox_annos

Iterate over all bbox annotation.

Returns:Iterator of lost.db.model.TwoDAnno.
data_exports

list of lost.pyapi.pipe_elements.VisualOutput objects.

datasources

list of lost.pyapi.pipe_elements.Datasource objects

img_annos

Iterate over all lost.db.model.ImageAnno objects in this Resultset.

Returns:Iterator of lost.db.model.ImageAnno objects.
line_annos

Iterate over all line annotations.

Returns:Iterator of lost.db.model.TwoDAnno objects.
mia_tasks

list of lost.pyapi.pipe_elements.MIATask objects

point_annos

Iterate over all point annotations.

Returns:Iterator of lost.db.model.TwoDAnno.
polygon_annos

Iterate over all polygon annotations.

Returns:Iterator of lost.db.model.TwoDAnno objects.
raw_files

list of lost.pyapi.pipe_elements.RawFile objects

sia_tasks

list of lost.pyapi.pipe_elements.SIATask objects

to_df()[source]

Get a pandas DataFrame of all annotations related to this object.

Returns:
Column names are:
’img.idx’, ‘img.anno_task_id’, ‘img.timestamp’, ‘img.timestamp_lock’, ‘img.state’, ‘img.sim_class’, ‘img.frame_n’, ‘img.video_path’, ‘img.img_path’, ‘img.result_id’, ‘img.iteration’, ‘img.group_id’, ‘img.anno_time’, ‘img.lbl.idx’, ‘img.lbl.name’, ‘img.lbl.external_id’, ‘img.annotator’, ‘anno.idx’, ‘anno.anno_task_id’, ‘anno.timestamp’, ‘anno.timestamp_lock’, ‘anno.state’, ‘anno.track_n’, ‘anno.dtype’, ‘anno.sim_class’, ‘anno.iteration’, ‘anno.group_id’, ‘anno.img_anno_id’, ‘anno.annotator’, ‘anno.confidence’, ‘anno.anno_time’, ‘anno.lbl.idx’, ‘anno.lbl.name’, ‘anno.lbl.external_id’, ‘anno.data’
Return type:pandas.DataFrame
to_vec(columns='all')[source]

Get a vector of all Annotations related to this object.

Parameters:columns (str or list of str) – ‘all’ OR ‘img.idx’, ‘img.anno_task_id’, ‘img.timestamp’, ‘img.timestamp_lock’, ‘img.state’, ‘img.sim_class’, ‘img.frame_n’, ‘img.video_path’, ‘img.img_path’, ‘img.result_id’, ‘img.iteration’, ‘img.group_id’, ‘img.anno_time’, ‘img.lbl.idx’, ‘img.lbl.name’, ‘img.lbl.external_id’, ‘img.annotator’, ‘anno.idx’, ‘anno.anno_task_id’, ‘anno.timestamp’, ‘anno.timestamp_lock’, ‘anno.state’, ‘anno.track_n’, ‘anno.dtype’, ‘anno.sim_class’, ‘anno.iteration’, ‘anno.group_id’, ‘anno.img_anno_id’, ‘anno.annotator’, ‘anno.confidence’, ‘anno.anno_time’, ‘anno.lbl.idx’, ‘anno.lbl.name’, ‘anno.lbl.external_id’, ‘anno.data’
Retruns:
list OR list of lists: Desired columns

Example

Return just a list of 2d anno labels:

>>> img_anno.to_vec('anno.lbl.name')
['Aeroplane', 'Bicycle', 'Bottle', 'Horse']

Return a list of lists:

>>> self.inp.get_anno_vec.(['img.img_path', 'anno.lbl.name',
...     'anno.data', 'anno.dtype'])
[
    ['path/to/img1.jpg', 'Aeroplane', [0.1, 0.1, 0.2, 0.2], 'bbox'],
    ['path/to/img1.jpg', 'Bicycle', [0.1, 0.1], 'point'],
    ['path/to/img2.jpg', 'Bottle', [[0.1, 0.1], [0.2, 0.2]], 'line'],
    ['path/to/img3.jpg', 'Horse', [0.2, 0.15, 0.3, 0.18], 'bbox']
]
twod_annos

Iterate over 2D-annotations.

Returns:of lost.db.model.TwoDAnno objects.
Return type:Iterator
visual_outputs

list of lost.pyapi.pipe_elements.VisualOutput objects.

pipeline

PipeInfo

class lost.pyapi.pipeline.PipeInfo(pipe, dbm)[source]
description

Description that was defined when pipeline was started.

Type:str
logfile_path

Path to pipeline log file.

Type:str
name

Name of this pipeline

Type:str
timestamp

Timestamp when pipeline was started.

Type:str
timestamp_finished

Timestamp when pipeline was finished.

Type:str

pipe_elements

Datasource

class lost.pyapi.pipe_elements.Datasource(pe, dbm)[source]
inp

Input of this pipeline element

Type:lost.pyapi.inout.Input
outp

Output of this pipeline element

Type:lost.pyapi.inout.Output
path

Absolute path to file or folder

Type:str
pipe_info

An object with pipeline informations

Type:lost.pyapi.pipeline.PipeInfo

RawFile

class lost.pyapi.pipe_elements.RawFile(pe, dbm)[source]
inp

Input of this pipeline element

Type:lost.pyapi.inout.Input
outp

Output of this pipeline element

Type:lost.pyapi.inout.Output
path

Absolute path to file or folder

Type:str
pipe_info

An object with pipeline informations

Type:lost.pyapi.pipeline.PipeInfo

AnnoTask

class lost.pyapi.pipe_elements.AnnoTask(pe, dbm)[source]
configuration

Configuration of this annotask.

Type:str
inp

Input of this pipeline element

Type:lost.pyapi.inout.Input
instructions

Instructions for the annotator of this AnnoTask.

Type:str
name

A name for this annotask.

Type:str
outp

Output of this pipeline element

Type:lost.pyapi.inout.Output
pipe_info

An object with pipeline informations

Type:lost.pyapi.pipeline.PipeInfo
possible_label_df

Get all possible labels for this annotation task in DataFrame format

pd.DataFrame: Column names are:
‘idx’, ‘name’, ‘abbreviation’, ‘description’, ‘timestamp’, ‘external_id’, ‘is_deleted’, ‘parent_leaf_id’ ,’is_root’
Type:pd.DataFrame
progress

Progress in percent.

Value range 0…100.

Type:float

MIATask

class lost.pyapi.pipe_elements.MIATask(pe, dbm)[source]
configuration

Configuration of this annotask.

Type:str
inp

Input of this pipeline element

Type:lost.pyapi.inout.Input
instructions

Instructions for the annotator of this AnnoTask.

Type:str
name

A name for this annotask.

Type:str
outp

Output of this pipeline element

Type:lost.pyapi.inout.Output
pipe_info

An object with pipeline informations

Type:lost.pyapi.pipeline.PipeInfo
possible_label_df

Get all possible labels for this annotation task in DataFrame format

pd.DataFrame: Column names are:
‘idx’, ‘name’, ‘abbreviation’, ‘description’, ‘timestamp’, ‘external_id’, ‘is_deleted’, ‘parent_leaf_id’ ,’is_root’
Type:pd.DataFrame
progress

Progress in percent.

Value range 0…100.

Type:float

SIATask

class lost.pyapi.pipe_elements.SIATask(pe, dbm)[source]
configuration

Configuration of this annotask.

Type:str
inp

Input of this pipeline element

Type:lost.pyapi.inout.Input
instructions

Instructions for the annotator of this AnnoTask.

Type:str
name

A name for this annotask.

Type:str
outp

Output of this pipeline element

Type:lost.pyapi.inout.Output
pipe_info

An object with pipeline informations

Type:lost.pyapi.pipeline.PipeInfo
possible_label_df

Get all possible labels for this annotation task in DataFrame format

pd.DataFrame: Column names are:
‘idx’, ‘name’, ‘abbreviation’, ‘description’, ‘timestamp’, ‘external_id’, ‘is_deleted’, ‘parent_leaf_id’ ,’is_root’
Type:pd.DataFrame
progress

Progress in percent.

Value range 0…100.

Type:float

DataExport

class lost.pyapi.pipe_elements.DataExport(pe, dbm)[source]
file_path

A list of absolute path to exported files

Type:list of str
inp

Input of this pipeline element

Type:lost.pyapi.inout.Input
outp

Output of this pipeline element

Type:lost.pyapi.inout.Output
pipe_info

An object with pipeline informations

Type:lost.pyapi.pipeline.PipeInfo
to_dict()[source]

Transform a list of exports to a dictionary.

Returns:[{‘iteration’:int, ‘file_path’:str},…]
Return type:list of dict

VisualOutput

class lost.pyapi.pipe_elements.VisualOutput(pe, dbm)[source]
html_strings

list of html strings.

Type:list of str
img_paths

List of absolute paths to images.

Type:list of str
inp

Input of this pipeline element

Type:lost.pyapi.inout.Input
outp

Output of this pipeline element

Type:lost.pyapi.inout.Output
pipe_info

An object with pipeline informations

Type:lost.pyapi.pipeline.PipeInfo
to_dict()[source]

Transforms a list of visualization information into a list of dicts.

Returns:[{‘iteration’:int, ‘img_path’:str, ‘html_string’:str},…]
Return type:list of dicts

Loop

class lost.pyapi.pipe_elements.Loop(pe, dbm)[source]
inp

Input of this pipeline element

Type:lost.pyapi.inout.Input
is_broken

True if loop is broken

Type:bool
iteration

Current iteration of this loop.

Type:int
max_iteration

Maximum number of iteration.

Type:int
outp

Output of this pipeline element

Type:lost.pyapi.inout.Output
pe_jump

PipelineElement where this loop will jump to when looping.

Can be of type:
pipe_info

An object with pipeline informations

Type:lost.pyapi.pipeline.PipeInfo

model

ImageAnno

class lost.db.model.ImageAnno(anno_task_id=None, user_id=None, timestamp=None, state=None, sim_class=None, result_id=None, img_path=None, frame_n=None, video_path=None, iteration=0, anno_time=None, is_junk=None, description=None)[source]

An ImageAnno represents an image annotation.

Multiple labels as well as 2d annotations (e.g. points, lines, boxes, polygons) can be assigned to an image.

labels

The related Label object.

Type:list
twod_annos

A list of TwoDAnno objects.

Type:list
img_path

Path to the image where this anno belongs to.

Type:str
frame_n

If this image is part of an video, frame_n indicates the frame number.

Type:int
video_path

If this image is part of an video, this should be the path to that video in file system.

Type:str
sim_class

The similarity class this anno belong to. It is used to cluster similar annos in MIA

Type:int
anno_time

Overall annotation time in seconds.

timestamp

Timestamp of ImageAnno

Type:DateTime
iteration

The iteration of a loop when this anno was created.

Type:int
idx

ID of this ImageAnno in database

Type:int
anno_task_id

ID of the anno_task this ImageAnno belongs to.

Type:int
state

See lost.db.state.Anno

Type:enum
result_id

Id of the related result.

user_id

Id of the annotator.

Type:int
is_junk

This image was marked as Junk.

Type:bool
description

Description for this annotation. Assigned by an annotator or algorithm.

Type:str
get_anno_vec(anno_type='bbox')[source]

Get related 2d annotations in list style.

Parameters:anno_type (str) – Can be ‘bbox’, ‘point’, ‘line’, ‘polygon’
Returns:
For POINTs:
[[x, y], [x, y], …]
For BBOXs:
[[x, y, w, h], [x, y, w, h], …]
For LINEs and POLYGONs:
[[[x, y], [x, y],…], [[x, y], [x, y],…]]
Return type:list of list of floats

Example

In the following example all bounding boxes of the image annotation will be returned in list style:

>>> img_anno.anno_vec()
[[0.1 , 0.2 , 0.3 , 0.18],
 [0.25, 0.25, 0.2, 0.4]]
>>> img_anno.get_anno_lbl_vec('name', 'bbox') #Get related label names
[['cow'], ['horse']]
iter_annos(anno_type='bbox')[source]

Iterator for all related 2D annotations of this image.

Parameters:anno_type (str) – Can be bbox’, ‘point’, ‘line’, ‘polygon’, ‘all’
Retruns:
iterator of TwoDAnno objects

Example

>>> for bb in img_anno.iter_annos('bbox'):
...     do_something(bb)
to_df()[source]

Tranform this ImageAnnotation and all related TwoDAnnotaitons into a pandas DataFrame.

Returns:
Column names are:
’img.idx’, ‘img.anno_task_id’, ‘img.timestamp’, ‘img.timestamp_lock’, ‘img.state’, ‘img.sim_class’, ‘img.frame_n’, ‘img.video_path’, ‘img.img_path’, ‘img.result_id’, ‘img.iteration’, ‘img.user_id’, ‘img.anno_time’, ‘img.lbl.idx’, ‘img.lbl.name’, ‘img.lbl.external_id’, ‘img.annotator’, ‘img.is_junk’, ‘anno.idx’, ‘anno.anno_task_id’, ‘anno.timestamp’, ‘anno.timestamp_lock’, ‘anno.state’, ‘anno.track_n’, ‘anno.dtype’, ‘anno.sim_class’, ‘anno.iteration’, ‘anno.user_id’, ‘anno.img_anno_id’, ‘anno.annotator’, ‘anno.confidence’, ‘anno.anno_time’, ‘anno.lbl.idx’, ‘anno.lbl.name’, ‘anno.lbl.external_id’, ‘anno.data’
Return type:pandas.DataFrame
to_dict(style='flat')[source]

Transform this ImageAnno and all related TwoDAnnos into a dict.

Parameters:style (str) – ‘flat’ or ‘hierarchical’. Return a dict in flat or nested style.
Returns:In ‘flat’ style return a list of dicts with one dict per annotation. In ‘hierarchical’ style, return a nested dictionary.
Return type:list of dict OR dict

Note

In ‘flat’ style annotation data and lists of labels are serialized as json strings. You may want to deserialize them with json.loads()

Example

HowTo iterate through all TwoDAnnotations of this ImageAnno dictionary in flat style:

>>> for d in img_anno.to_dict():
...     print(d['img.img_path'], d['anno.lbl.name'], d['anno.dtype'])
path/to/img1.jpg ['Aeroplane'] bbox
path/to/img1.jpg ['Bicycle'] point

Possible keys in flat style:

>>> img_anno.to_dict()[0].keys()
dict_keys([
    'img.idx', 'img.anno_task_id', 'img.timestamp',
    'img.timestamp_lock', 'img.state', 'img.sim_class',
    'img.frame_n', 'img.video_path', 'img.img_path',
    'img.result_id', 'img.iteration', 'img.user_id',
    'img.anno_time', 'img.lbl.idx', 'img.lbl.name',
    'img.lbl.external_id', 'img.annotator', 'img.is_junk'
    'anno.idx', 'anno.anno_task_id', 'anno.timestamp',
    'anno.timestamp_lock', 'anno.state', 'anno.track_n',
    'anno.dtype', 'anno.sim_class', 'anno.iteration',
    'anno.user_id', 'anno.img_anno_id', 'anno.annotator',
    'anno.confidence', 'anno.anno_time', 'anno.lbl.idx',
    'anno.lbl.name', 'anno.lbl.external_id', 'anno.data'
])

HowTo iterate through all TwoDAnnotations of this ImageAnno dictionary in hierarchical style:

>>> h_dict = img_anno.to_dict(style='hierarchical')
>>> for d in h_dict['img.twod_annos']:
...     print(h_dict['img.img_path'], d['anno.lbl.name'], d['anno.dtype'])
path/to/img1.jpg [Aeroplane] bbox
path/to/img1.jpg [Bicycle] point

Possible keys in hierarchical style:

>>> h_dict = img_anno.to_dict(style='hierarchical')
>>> h_dict.keys()
dict_keys([
    'img.idx', 'img.anno_task_id', 'img.timestamp',
    'img.timestamp_lock', 'img.state', 'img.sim_class',
    'img.frame_n', 'img.video_path', 'img.img_path',
    'img.result_id', 'img.iteration', 'img.user_id',
    'img.anno_time', 'img.lbl.idx', 'img.lbl.name',
    'img.lbl.external_id', 'img.annotator', 'img.twod_annos'
])
>>> h_dict['img.twod_annos'][0].keys()
dict_keys([
    'anno.idx', 'anno.anno_task_id', 'anno.timestamp',
    'anno.timestamp_lock', 'anno.state', 'anno.track_n',
    'anno.dtype', 'anno.sim_class', 'anno.iteration',
    'anno.user_id', 'anno.img_anno_id', 'anno.annotator',
    'anno.confidence', 'anno.anno_time', 'anno.lbl.idx',
    'anno.lbl.name', 'anno.lbl.external_id', 'anno.data'
])
to_vec(columns='all')[source]

Transform this ImageAnnotation and all related TwoDAnnotations in list style.

Parameters:columns (str or list of str) – ‘all’ OR ‘img.idx’, ‘img.anno_task_id’, ‘img.timestamp’, ‘img.timestamp_lock’, ‘img.state’, ‘img.sim_class’, ‘img.frame_n’, ‘img.video_path’, ‘img.img_path’, ‘img.result_id’, ‘img.iteration’, ‘img.user_id’, ‘img.anno_time’, ‘img.lbl.idx’, ‘img.lbl.name’, ‘img.lbl.external_id’, ‘img.annotator’, ‘img.is_junk’, ‘anno.idx’, ‘anno.anno_task_id’, ‘anno.timestamp’, ‘anno.timestamp_lock’, ‘anno.state’, ‘anno.track_n’, ‘anno.dtype’, ‘anno.sim_class’, ‘anno.iteration’, ‘anno.user_id’, ‘anno.img_anno_id’, ‘anno.annotator’, ‘anno.confidence’, ‘anno.anno_time’, ‘anno.lbl.idx’, ‘anno.lbl.name’, ‘anno.lbl.external_id’, ‘anno.data’
Retruns:
list OR list of lists: Desired columns

Example

Return just a list of serialized 2d anno labels:

>>> img_anno.to_vec('anno.lbl.name')
["['Aeroplane']", "['Bicycle']"]

Return a list of lists:

>>> img_anno.to_vec(['img.img_path', 'anno.lbl.name',
...     'anno.lbl.idx', 'anno.dtype'])
[
    ['path/to/img1.jpg', "['Aeroplane']", 14, 'bbox'],
    ['path/to/img1.jpg', "['Bicycle']", 15, 'point']
]

TwoDAnno

class lost.db.model.TwoDAnno(anno_task_id=None, user_id=None, timestamp=None, state=None, track_id=None, sim_class=None, img_anno_id=None, timestamp_lock=None, iteration=0, data=None, dtype=None, confidence=None, anno_time=None, description=None)[source]

A TwoDAnno represents a 2D annotation/ drawing for an image.

A TwoDAnno can be of type point, line, bbox or polygon.

idx

ID of this TwoDAnno in database

Type:int
anno_task_id

ID of the anno_task this TwoDAnno belongs to.

Type:int
timestamp

Timestamp created of TwoDAnno

Type:DateTime
timestamp_lock

Timestamp locked in view

Type:DateTime
state

can be unlocked, locked, locked_priority or labeled (see lost.db.state.Anno)

Type:enum
track_id

The track id this TwoDAnno belongs to.

Type:int
sim_class

The similarity class this anno belong to. It is used to cluster similar annos in MIA.

Type:int
iteration

The iteration of a loop when this anno was created.

Type:int
user_id

Id of the annotator.

Type:int
img_anno_id

ID of ImageAnno this TwoDAnno is appended to

Type:int
data

drawing data (for e.g. x,y, width, height) of anno - depends on dtype

Type:Text
dtype

type of TwoDAnno (for e.g. bbox, polygon) (see lost.db.dtype.TwoDAnno)

Type:int
labels

A list of Label objects related to the TwoDAnno.

Type:list
confidence

Confidence of Annotation.

Type:float
anno_time

Overall Annotation Time in ms.

description

Description for this annotation. Assigned by an annotator or algorithm.

Type:str
add_label(label_leaf_id)[source]

Add a label to this 2D annotation.

Parameters:label_leaf_id (int) – Id of the label_leaf that should be added.
bbox

BBOX annotation in list style [x, y, w, h]

Example

>>> anno = TwoDAnno()
>>> anno.bbox = [0.1, 0.1, 0.2, 0.2]
>>> anno.bbox
[0.1, 0.1, 0.2, 0.2]
Type:list
get_anno_vec()[source]

Get annotation data in list style.

Returns:
For a POINT:
[x, y]
For a BBOX:
[x, y, w, h]
For a LINE and POLYGONS:
[[x, y], [x, y],…]
Return type:list of floats

Example

HowTo get a numpy array? In the following example a bounding box is returned:

>>> np.array(twod_anno.get_anno_vec())
array([0.1 , 0.2 , 0.3 , 0.18])
line

LINE annotation in list style [[x, y], [x, y], …]

Example

>>> anno = TwoDAnno()
>>> anno.line = [[0.1, 0.1], [0.2, 0.2]]
>>> anno.line
[[0.1, 0.1], [0.2, 0.2]]
Type:list of list
point

POINT annotation in list style [x, y]

Example

>>> anno = TwoDAnno()
>>> anno.point = [0.1, 0.1]
>>> anno.point
[0.1, 0.1]
Type:list
polygon

polygon annotation in list style [[x, y], [x, y], …]

Example

>>> anno = TwoDAnno()
>>> anno.polygon = [[0.1, 0.1], [0.2, 0.1], [0.15, 0.2]]
>>> anno.polygon
[[0.1, 0.1], [0.2, 0.1], [0.15, 0.2]]
Type:list of list
to_df()[source]

Transform this annotation into a pandas DataFrame

Returns:A DataFrame where column names correspond to the keys of the dictionary returned from to_dict() method.
Return type:pandas.DataFrame

Note

Column names are:
[‘anno.idx’, ‘anno.anno_task_id’, ‘anno.timestamp’, ‘anno.timestamp_lock’, ‘anno.state’, ‘anno.track_id’, ‘anno.dtype’, ‘anno.sim_class’, ‘anno.iteration’, ‘anno.user_id’, ‘anno.img_anno_id’, ‘anno.annotator’, ‘anno.confidence’, ‘anno.anno_time’, ‘anno.lbl.idx’, ‘anno.lbl.name’, ‘anno.lbl.external_id’, ‘anno.data’]
to_dict(style='flat')[source]

Transform this object into a dict.

Parameters:style (str) – ‘flat’ or ‘hierarchical’ ‘flat’: Return a dictionray in table style ‘hierarchical’: Return a nested dictionary
Retruns:
dict: In flat or hierarchical style.

Example

Get a dict in flat style. Note that ‘anno.data’, ‘anno.lbl.idx’, ‘anno.lbl.name’ and ‘anno.lbl.external_id’ are json strings in contrast to the hierarchical style.

>>> bbox.to_dict(style='flat')
{
    'anno.idx': 88,
    'anno.anno_task_id': None,
    'anno.timestamp': None,
    'anno.timestamp_lock': None,
    'anno.state': None,
    'anno.track_id': None,
    'anno.dtype': 'bbox',
    'anno.sim_class': None,
    'anno.iteration': 0,
    'anno.user_id': 47,
    'anno.img_anno_id': None,
    'anno.annotator': 'test',
    'anno.confidence': None,
    'anno.anno_time': None,
    'anno.lbl.idx': '["14"]',
    'anno.lbl.name': '["Aeroplane"]',
    'anno.lbl.external_id': '["6"]',
    'anno.data': '{"x": 0.1, "y": 0.1, "w": 0.2, "h": 0.2}'
}

Get a dict in hierarchical style. Note that ‘anno.data’ is a dict in contrast to the flat style.

>>> bbox.to_dict(style='hierarchical')
{
    'anno.idx': 86,
    'anno.anno_task_id': None,
    'anno.timestamp': None,
    'anno.timestamp_lock': None,
    'anno.state': None,
    'anno.track_id': None,
    'anno.dtype': 'bbox',
    'anno.sim_class': None,
    'anno.iteration': 0,
    'anno.user_id': 46,
    'anno.img_anno_id': None,
    'anno.annotator': 'test',
    'anno.confidence': None,
    'anno.anno_time': None,
    'anno.lbl.idx': [14],
    'anno.lbl.name': ['Aeroplane'],
    'anno.lbl.external_id': ['6'],
    'anno.data': {
        'x': 0.1, 'y': 0.1, 'w': 0.2, 'h': 0.2
    }
}
to_vec(columns='all')[source]

Tansfrom this annotation in list style.

Parameters:columns (list of str OR str) – Possible column names are: ‘all’ OR [‘anno.idx’, ‘anno.anno_task_id’, ‘anno.timestamp’, ‘anno.timestamp_lock’, ‘anno.state’, ‘anno.track_id’, ‘anno.dtype’, ‘anno.sim_class’, ‘anno.iteration’, ‘anno.user_id’, ‘anno.img_anno_id’, ‘anno.annotator’, ‘anno.confidence’, ‘anno.anno_time’, ‘anno.lbl.idx’, ‘anno.lbl.name’, ‘anno.lbl.external_id’, ‘anno.data’]
Returns:A list of the desired columns.
Return type:list of objects

Example

If you want to get only the annotation in list style e.g. [x, y, w, h] (if this TwoDAnnotation is a bbox).

>>> anno.to_vec('anno.data')
[0.1, 0.1, 0.2, 0.2]

If you want in addition also the corresponding label names and label ids for this annotation then just add additional column names:

>>> bbox.to_vec(['anno.data', 'anno.lbl.idx', 'anno.lbl.name'])
[[0.1, 0.1, 0.2, 0.2], "[14]", "['Aeroplane']"]

LabelLeaf

class lost.db.model.LabelLeaf(idx=None, name=None, abbreviation=None, description=None, timestamp=None, external_id=None, label_tree_id=None, is_deleted=None, parent_leaf_id=None, is_root=None)[source]

A LabelLeaf

idx

ID in database.

Type:int
name

Name of the LabelName.

Type:str
abbreviation
Type:str
description
Type:str
timestamp
Type:DateTime
external_id

Id of an external semantic label system (for e.g. synsetid of wordnet)

Type:str
is_deleted
Type:Boolean
is_root

Indicates if this leaf is the root of a tree.

Type:Boolean
parent_leaf_id

Reference to parent LabelLeaf.

Type:Integer
label_leafs
Type:list of LabelLeaf
to_df()[source]

Transform this LabelLeaf to a pandas DataFrame.

Returns:
Return type:pd.DataFrame
to_dict()[source]

Transform this object to a dict.

Returns:
Return type:dict

Label

class lost.db.model.Label(idx=None, dtype=None, label_leaf_id=None, img_anno_id=None, two_d_anno_id=None, annotator_id=None, timestamp_lock=None, timestamp=None, confidence=None, anno_time=None)[source]

Represants an Label that is related to an annoation.

idx

ID in database.

Type:int
dtype

lost.db.dtype.Result type of this attribute.

Type:enum
label_leaf_id

ID of related model.LabelLeaf.

img_anno_id
Type:int
two_d_anno_id
Type:int
timestamp
Type:DateTime
timestamp_lock
Type:DateTime
label_leaf

related model.LabelLeaf object.

Type:model.LabelLeaf
annotator_id

GroupID of Annotator who has assigned this Label.

Type:Integer
confidence

Confidence of Annotation.

Type:float
anno_time

Time of annotaiton duration

Type:float

logic.label

LabelTree

class lost.logic.label.LabelTree(dbm, root_id=None, root_leaf=None, name=None, logger=None)[source]

A class that represants a LabelTree.

Parameters:
  • dbm (lost.db.access.DBMan) – Database manager object.
  • root_id (int) – label_leaf_id of the root Leaf.
  • root_leaf (lost.db.model.LabelLeaf) – Root leaf of the tree.
  • name (str) – Name of a label tree.
  • logger (logger) – A logger.
create_child(parent_id, name, external_id=None)[source]

Create a new leaf in label tree.

Parameters:
  • parent_id (int) – Id of the parend leaf.
  • name (str) – Name of the leaf e.g the class name.
  • external_id (str) – Some id of an external label system.
Retruns:
lost.db.model.LabelLeaf: The the created child leaf.
create_root(name, external_id=None)[source]

Create the root of a label tree.

Parameters:
  • name (str) – Name of the root leaf.
  • external_id (str) – Some id of an external label system.
Retruns:
lost.db.model.LabelLeaf or None:
The created root leaf or None if a root leaf with same name is already present in database.
delete_subtree(leaf)[source]

Recursive delete all leafs in subtree starting with leaf

Parameters:leaf (lost.db.model.LabelLeaf) – Delete all childs of this leaf. The leaf itself stays.
delete_tree()[source]

Delete whole tree from system

get_child_vec(parent_id, columns='idx')[source]

Get a vector of child labels.

Parameters:
  • parent_id (int) – Id of the parent leaf.
  • columns (str or list of str) – Can be any attribute of lost.db.model.LabelLeaf for example ‘idx’, ‘external_idx’, ‘name’ or a list of these e.g. [‘name’, ‘idx’]

Example

>>> label_tree.get_child_vec(1, columns='idx')
[2, 3, 4]
>>> label_tree.get_child_vec(1, columns=['idx', 'name'])
[
    [2, 'cow'],
    [3, 'horse'],
    [4, 'person']
]
Returns:
Return type:list in the requested columns
import_df(df)[source]

Import LabelTree from DataFrame

Parameters:df (pandas.DataFrame) – LabelTree in DataFrame style.
Retruns:
lost.db.model.LabelLeaf or None:
The created root leaf or None if a root leaf with same name is already present in database.
to_df()[source]

Transform this LabelTree to a pandas DataFrame.

Returns:pandas.DataFrame

dtype

TwoDAnno

class lost.db.dtype.TwoDAnno[source]

Type of a TwoDAnno

BBOX

A BBox.

Type:1
POLYGON

A Polygon.

Type:2
POINT

A Point.

Type:3
LINE

A Line.

Type:4
CIRCLE

A Circle.

Type:5

util methods

anno_helper

A module with helper methods to tranform annotations into different formats and to crop annotations from an image.

lost.pyapi.utils.anno_helper.calc_box_for_anno(annos, types, point_padding=0.05)[source]

Calculate a bouning box for an arbitrary 2DAnnotation.

Parameters:
  • annos (list) – List of annotations.
  • types (list) – List of types.
  • point_padding (float, optional) – In case of a point we need to add some padding to get a box.
Returns:

A list of bounding boxes in format [[xc,yc,w,h],…]

Return type:

list

lost.pyapi.utils.anno_helper.crop_boxes(annos, types, img, context=0.0, draw_annotations=False)[source]

Crop a bounding boxes for TwoDAnnos from image.

Parameters:
  • annos (list) – List of annotations.
  • types (list) – List of types.
  • img (numpy.array) – The image where boxes should be cropped from.
  • context (float) – The context that should be added to the box.
  • draw_annotations (bool) – If true, annotation will be painted inside the crop.
Returns:

A tuple that contains a list of image crops and a list of bboxes [[xc,yc,w,h],…]

Return type:

(list of numpy.array, list of list of float)

lost.pyapi.utils.anno_helper.divide_into_patches(img, x_splits=2, y_splits=2)[source]

Divide image into x_splits*y_splits patches.

Parameters:
  • img (array) – RGB image (skimage.io.imread).
  • x_splits (int) – Number of elements on x axis.
  • y_splits (int) – Number of elements on y axis.
Returns:

img_patches, box_coordinates

img batches and box coordinates of these patches in the image.

Return type:

list, list

Note

img_patches are in following order:
[[x0,y0], [x0,y1],…[x0,yn],…,[xn,y0], [xn, y1]…[xn,yn]]
lost.pyapi.utils.anno_helper.draw_annos(annos, types, img, color=(255, 0, 0), point_r=2)[source]

Draw annotations inside a image

Parameters:
  • annos (list) – List of annotations.
  • types (list) – List of types.
  • img (numpy.array) – The image to draw annotations in.
  • color (tuple) – (R,G,B) color that is used for drawing.

Note

The given image will be directly edited!

Returns:Image with drawn annotations
Return type:numpy.array
lost.pyapi.utils.anno_helper.to_abs(annos, types, img_size)[source]

Convert relative annotation coordinates to absolute ones

Parameters:
  • annos (list of list) –
  • types (list of str) –
  • img_size (tuple) – (width, height) of the image in pixels.
Returns:

Annotations in absolute format.

Return type:

list of list

lost.pyapi.utils.anno_helper.trans_boxes_to(boxes, convert_to='minmax')[source]

Transform a box from standard lost format into a different format

Parameters:
  • boxes (list of list) – Boxes in standard lost format [[xc,yc,w,h],…]
  • convert_to (str) – minmax -> [[xmim,ymin,xmax,ymax]…]
Returns:

Converted boxes.

Return type:

list of list

blacklist

A helper module to deal with blacklists.

class lost.pyapi.utils.blacklist.ImgBlacklist(my_script, name='img-blacklist.json', context='pipe')[source]

A class to deal with image blacklists.

Such blacklists are often used for annotation loops, in order to prevent annotating the same image multiple times.

my_script

The script instance that creates this blacklist.

Type:lost.pyapi.script.Script
name

The name of the blacklist file.

Type:str
context

Options: instance, pipe, static

Type:str

Example

Add images to blacklist.

>>> blacklist = ImgBlacklist(self, name='blacklist.json')
>>> blacklist.add('path/to/img0.jpg')
>>> balcklist.save()

Load a blacklist and check if a certain image is already in list.

>>> blacklist = ImgBlacklist(self, name='blacklist.json')
>>> blacklist.contains('path/to/img0.jpg')
True
>>> blacklist.contains('path/to/img1.jpg')
False

Get list of images that are not part of the blacklist

>>> blacklist.get_whitelist(['path/to/img0.jpg', 'path/to/img1.jpg', 'path/to/img2.jpg'])
['path/to/img1.jpg', 'path/to/img2.jpg']

Add images to the blacklist

>>> blacklist.add(['path/to/img1.jpg', 'path/to/img2.jpg'])
add(imgs)[source]

Add an image to the blacklist.

Parameters:imgs (list) – A list of image identifiers that should be added to the blacklist.
contains(img)[source]

Check if blacklist contains a spcific image

Parameters:img (str) – The image identifier
Returns:True if img in blacklist, False if not.
Return type:bool
delete_blacklist()[source]

Remove blacklist from filesystem

get_whitelist(img_list, n='all')[source]

Get a list of images that are not part of the blacklist.

Parameters:
  • img_list (list of str) – A list of images where should be checked if they are in the blacklist
  • n ('all' or 'int') – The maximum number of images that should be returned.
Returns:

A list of images that are not in the blacklist.

Return type:

list of str

remove_item(item)[source]

Remove item from blacklist

Parameters:item (str) – The item/ image to remove from blacklist.
save()[source]

Write blacklist to filesystem

vis

lost.pyapi.utils.vis.boxes(script, img_anno, figsize=(15, 15), fontsize=15, label_offset=(0, 15))[source]

Draw bboxes on into an matplotlib figures

Parameters:
  • script (lost.pyapi.script.Script) – The script object that uses this method.
  • img_anno (lost.pyapi.annos.ImageAnno) – The image anno where bboxes should be visualized for.
  • figsize (tuple) – Size of the matplotlib figure
  • fontsize (ing) – Fontsize in pixels for label display
  • label_offset (tuple) – Position of the label in pixels in relation to the upper left corner of the box.
Returns:

Matplotlib figure

lost.pyapi.utils.vis.vis_tracks(img, tracks, frame_n, dots=15, figsize=(10, 10), dot_radius=5, linewidth=2)[source]

Visualize a track on image

Parameters:
  • img (array or str) – An RGB image or path to the image file.
  • tracks (array) – [[frame_n, track_id, xc, yc, w, h]…[…]] Box is defined relative to the image.
  • frame_n (int) – The frame number belonning to the image
  • dots (int) – Number of dots that will be displayed. Past locations that will be visualized.
  • figsize (tuple) – (int,int) Size of the figure to display.
  • dot_radius (int) – Radius of the first dot.
  • linewidth (int) – Linewidth of the box to draw.
Returns:

Matplotlib figure