recognize 0.1.21

CLI for ML Pipeline

Recognizes

This CLI allows you to upload files to the ML Pipeline and query results.

First Steps

You first need to make sure that you have the requirements needed to install the CLI as well as install the CLI itself.

Make sure that you have python and pip installed. It is recommended that you use a virtual environment although not a requirement.

Then to install the CLI run the following command in your shell:

pip install recognize

Usage

After installing the CLI run the following command:

recognize --help

This should show you a list of all the available commands with a short description of what they do. You can append the --help option to each command for more details about the command. For example:

recognize upload --help

This should give you a list of all the subcommands for this command, the options, and the order in which to write them.

Examples

To upload a directory of videos, accepted file types are mp4 and mov:

recognize upload directory path/to/directory/of/videos --tag some_custom_tag

The tag is useful for if you want to organise the uploads. You can search for labels associated with the tag and you can upload files using this tag at a later stage, and they will be automatically grouped.

After uploading videos it will take a few minutes to process them all. After which you can start querying.

recognize search keywords --output results.csv some interesting words

The --output argument is optional, if ommited you will be asked what to name the output interactively instead.

You can also search for specific entry types returned by AWS Rekognition. For example to which resources might have potentially harmful, violent, or offensive content run the following command:

recognize search entries --output results.csv moderation

Finally, to search for a particular face run the following command, accepted file types are png and jpeg:

recognize search faces --output results.csv file/path/to/image/of/face

Results

For keyword and entry search the columns in the csv (or keys in the records for the json output) will be:

Column	Description
id	The unique identifier for the entry.
average_confidence	The average confidence score.
confidences	The list of confidence scores.
entry_type	The type of entry. Either label, alias, category, parent, moderation.
name	The value associated with the entry.
timestamps	The timestamps related to the entry.
url	The S3 bucket URL of the associated video.
tag	The tag related to the entry.

The faces search will have the following columns (or keys):

Column	Description
faceId	The unique identifier for the face.
imageId	The unique identifier for the image.
externalImageId	The path to the resource in the S3 bucket with metadata.
similarity	The similarity of the face to the input image.
searchedFaceConfidence	Confidence in whether a face was detected in the input.
timestamp	The timestamp of the face. In milliseconds.