imgtools-m8 0.0.3

Simple image tools package. Used to convert, downscale or upscale images.

imgtools_m8

Simple image tools package. Used to convert, downscale and/or upscale images.

Use deep learning and cv2 to upscale image using Xavier Weber models, (more info here).

Installation

Install from GitHub repository :

To install directly from GitHub:

$ python3 -m pip install "git+https://github.com/mano8/imgtools_m8 --upgrade"

To install from PypI :

python3 -m pip install imgtools_m8 --upgrade

How to use

This package automatically covert, downscale and/or upscale an image file or a list of images from directory defined in source_path property.
(See accepted extensions from cv2 documentation)

Example :

    >>> # Set source as an image file 
    >>> source_path = /path/to/image.png
    >>> # Or set source as a directory containing your images
    >>> source_path = /path/to/directory

Next you will need to define output configuration:

• path: The output path
• output_formats: The list of output formats,
  who contains optional output sizes and output image format
  (extension and compression options)

It is possible to resize images with different options:

• fixed_width: resize image to exact width (in pixel)
• fixed_height: resize image to exact height (in pixel)
• fixed_size: resize image depending on first limitation reached (height or width)
• fixed_width and fixed_height: resize image depending on first limitation reached.
  same as fixed_size but can set different height/width values.

Example :
The source file is 340px width and 216px height.

Recien llegado by @Cezar yañez

We want resized output file to exact width of 1900px and 1200px. And we need output formats as JPEG (with 80% quality) and WEBP (with 70% quality)

    >>> source_path = "./tests/dummy_dir/recien_llegado.jpg"
    >>> output_conf = {
            # Output path
            'path': /my/output/path/directory, 
            'output_formats': [
                {  # Get resized output file to exact width of 1600px
                    'fixed_width': 1900,
                    'formats': [
                        # JPEg defauts are 'quality': 95, 'progressive': 0, 'optimize': 0
                        {'ext': '.jpg', 'quality': 80, 'progressive': 1, 'optimize': 1},
                    ]
                },
                {  # Get resized output file to exact width of 800px
                    'fixed_width': 1200,
                    'formats': [
                        {'ext': '.jpg', 'quality': 80},
                        {'ext': '.webp', 'quality': 70}
                        {'ext': '.png', 'compression': 2}
                    ]
                }
            ]

        }
        >>> imgtools = ImageTools(
            source_path=source_path,
            output_conf=output_conf
        )
        >>> imgtools.run()

This will create 4 files in the output directory :

• Two JPEG files resized as defined width (1900px and 1200px), with 80% quality, JPEG progressive and optimize features enabled
• Two WEBP files resized as defined width (1900px and 1200px), with 70% quality
• One PNG file resized as defined width (1900px), with PNG compression level = 2

The output file names are set as:

    >>> 'originalName'_'imageWidth'x'imageHeight'.'outputExtension'
    >>> # egg :
    >>> originalFileName_1400x1360.jpeg

One of above results is :

recien_llegado_1200x762.jpg by @Cezar yañez

In this case source file is precessed as:

• upscale 2x (source file is now 680px/432px)
• upscale 2x (source file is now 1360px/864px)
• for 1200px width output downscale and save as recien_llegado_1200x762.jpg recien_llegado_1200x762.webp recien_llegado_1200x762.png
• upscale 2x (source file is now 2720px/1728px)
• for 1900px width output downscale and save as recien_llegado_1900x1207.jpeg an .webp

By default, the image tool use EDSR_x2.pb deep learning model, to improve quality.

To load any compatible model of your choice to upscale the images, you can define model_conf property.

The imgtools_m8 only contain EDSR (x2, x3, x4) and FSRCNN(x2, x3, x4) models.

If you want uses one of them set model_conf property as:

    >>> # Set EDSR_x4 model to upscale images
    >>> model_conf = {
        'file_name': 'EDSR_x4.pb',
    }
    >>> # Or set FSRCNN_x2 model to upscale images
    >>> model_conf = {
        'file_name': 'FSRCNN_x2.pb',
    }

In case you want uses another compatible model, you may download the desired .pb file, and set model_conf property as:

    >>> # Set TF-ESPCN_x2 model to upscale images
    >>> model_conf = {
        'path': "/path/to/your/downloaded/model/directory",
        'file_name': 'TF-ESPCN_x2.pb',
    }

Here a complete example using TF-ESPCN_x2 model to upscale images :

    >>> # Set TF-ESPCN_x2 model to upscale images
    >>> model_conf = {
        'path': "/path/to/your/downloaded/model/directory",
        'file_name': 'TF-ESPCN_x2.pb',
    }
    >>> source_path = "./tests/dummy_dir"
    >>> output_conf = {
            # Output path
            'path': /my/output/path/directory, 
            'output_formats': [
                {  # Get resized output file to exact width of 1600px
                    'fixed_width': 1900,
                    'formats': [
                        {'ext': '.jpg', 'quality': 80},
                        {'ext': '.webp', 'quality': 70}
                    ]
                },
                {  # Get resized output file to exact width of 800px
                    'fixed_width': 1200,
                    'formats': [
                        {'ext': '.jpg', 'quality': 80},
                        {'ext': '.webp', 'quality': 70}
                    ]
                }
            ]

        }
        >>> imgtools = ImageTools(
            source_path=source_path,
            output_conf=output_conf,
            model_conf=model_conf
        )
        >>> imgtools.run()