libimgsmlr 0.1.6

Implement the feature extraction function of the ``imgsmlr`` extension for ``PostgreSQL``.

libimgsmlr

Implement the feature extraction function of the imgsmlr extension, in languages such as: c(so file), Python. Other languages should probably call dynamic libraries to adapt.

When using the PostgreSQL extension imgsmlr, I found that the feature extraction functions provided by imgsmlr are all PostgreSQL functions. This poses a problem, as the files need to be uploaded to and stored in the database during testing. This is generally not possible in service development, where text is usually stored separately in the file system, and only paths and pattern, signature are stored in the database. Therefore, it is necessary to implement the feature extraction functions used by the service development language.

Depends

This library depends on the gd library. Please install it yourself first.

On Mac OSX: brew install gd

On Linux: yum install -y gd-devel

Installation

pip install --global-option=build_ext libimgsmlr

Errors

If the following error is encountered on Mac OSX.

fatal error: 'gd.h' file not found
#include "gd.h"

The solution is to set the environment variable first:

export CFLAGS="-I$(brew --prefix gd)/include"
export LFDLAGS="-L$(brew --prefix gd)/lib"

Examples

import argparse
from libimgsmlr import img2pattern, pattern2signature, shuffle_pattern, pattern_distance, signature_distance


def image_pattern_signature(content):
    pattern = img2pattern(content)
    signature = pattern2signature(pattern)
    pattern = shuffle_pattern(pattern)
    return pattern, signature

def process_single_image(filename):
    content = open(filename, "rb").read()
    try:
        pattern, signature = image_pattern_signature(content)
        print(f"pattern: {pattern.as_array()}")
        print(f"signature: {signature}")
    except ValueError as err:
        print(f"Error: img2pattern({filename}) failed! {err}")

def calculate_distance(filename1, filename2):
    try:
        content1 = open(filename1, "rb").read()
        content2 = open(filename2, "rb").read()
        pattern1, signature1 = image_pattern_signature(content1)
        pattern2, signature2 = image_pattern_signature(content2)
        pattern_dist = pattern_distance(pattern1, pattern2)
        print(f"Pattern distance: {pattern_dist}")

        sig_dist = signature_distance(signature1, signature2)
        print(f"Signature distance: {sig_dist}")
    except ValueError as err:
        print(f"Error: Failed to calculate distance! {err}")

if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="libimgsmlr test")
    parser.add_argument('filenames', nargs='+', help='Image files (jpeg/png/gif)')
    args = parser.parse_args()
    if len(args.filenames) == 1:
        process_single_image(args.filenames[0])
    elif len(args.filenames) == 2:
        calculate_distance(args.filenames[0], args.filenames[1])
    else:
        print("Please provide 1 or 2 image filenames")

LICENSE

MIT