Results of tagging
TaggingResult
Once tagging is done you get a TaggingResult.
TaggingResult contains results of tagging for a single medium and contains the following elements:
processed_at- time when medium was processed.identifier- unique identifier of medium. For file based media - base path without extension, for YouTube video - video_id, for texts - text itself.type- type of medium.content- result of tagging. Can be either array of Tag or one or many Descriptions.tagger- type of tagger used for tagging this medium.output- type of operation on medium - either tag or description.tagging_details- Any details used to perform the tagging, for example custom_prompt, number of tags requested.
from media_tagging import tagging_result
result = tagging_result.TaggingResult(
tagger='gemini',
type='IMAGE',
output='tag',
content=[
tagging_result.Tag(name='tag_name_1', score=1.0),
tagging_result.Tag(name='tag_name_1', score=0.1),
],
tagging_details={'n_tags': 2},
from media_tagging import tagging_result
result = tagging_result.TaggingResult(
tagger='gemini',
type='IMAGE',
output='tag',
content= tagging_result.Description(text='True'),
tagging_details={
'custom_prompt': 'Is this an advertising. Answer only True or False'
},
from media_tagging import tagging_result
result = tagging_result.TaggingResult(
tagger='gemini',
type='IMAGE',
output='tag',
content=
tagging_result.Description(text='00:00:05'),
tagging_result.Description(text='00:00:15'),
],
tagging_details={
'custom_prompt': 'Find all the timestamps where a cat appears'
},
You can extract tags / descriptions themself via .content property
tags = results.content
Tag
Tag represents a unique concept (using a singular noun) with a score from 0 to 1 where 0 is an absence of a concept and 1 is a total presence.
Usually a single media is tagged with multiple tags with varying scores.
from media_tagging import tagging_result
tag = tagging_result.Tag(name='tag_name_1', score=0.1)
Trimming tags
If TaggingResult contains tags you can trim them by removing those which scores lower than a custom threshold. This can be useful when media-tagger provides a lot of non-definitive tags.
result.trim_tags(0.5)
Description
Description contains an information on a media in a form of a text or JSON.
from media_tagging import tagging_result
description = tagging_result.Description(text='This is an advertising.')