image-cropping-python - Transformation Block - Batch Crop, Video Frame Extraction, and Save Images

This Transformation block allows you to batch crop multiple images and extract/crop frames from videos in a folder to a specified size, saving the cropped versions to a new folder. The script can be used in a Jupyter Notebook or as a command-line tool. For videos, frames are extracted at a user-defined interval and cropped just like images. This is useful for image preprocessing tasks like dataset preparation, video dataset creation, image resizing for thumbnails, and other image manipulation tasks.

Features

• Batch processing of images and videos from a folder.
• Crops images to a user-defined size.
• For videos, extracts frames at a user-defined interval (in seconds), crops, and saves as images.
• Saves the cropped images in a specified output folder.
• Displays a preview of the first few cropped images for quick verification (notebook only).

Requirements

Before running the notebook or script, make sure you have the following libraries installed:

• opencv-python (for video frame extraction)
• Pillow (for image manipulation)
• matplotlib (for displaying the cropped image samples)

You can install the necessary libraries using pip:

pip install opencv-python pillow matplotlib

Usage in Jupyter notebook

1. Clone or download the repository

git clone [email protected]:moe-sani/image-cropping-python.git
cd image-cropper

2. Open the Jupyter Notebook

Open the provided Jupyter Notebook (Image_Cropper.ipynb) in your local environment.

3. Set Parameters

In the notebook, you'll need to set the following parameters:

• input_folder: Path to the folder containing the images you want to crop.
• output_folder: Path to the folder where cropped images will be saved.
• crop_size: Tuple specifying the size of the crop in (width, height) format.
• crop_x (optional): X coordinate for the top-left corner of the crop box. If not set, cropping is centered horizontally.
• crop_y (optional): Y coordinate for the top-left corner of the crop box. If not set, cropping is centered vertically.

For example:

input_folder = 'path/to/your/input/folder'
output_folder = 'path/to/your/output/folder'
crop_size = (300, 300)
crop_x = 50  # Optional
crop_y = 100 # Optional

4. Run the Notebook

Execute the notebook cells. The script will:

1. Load images from the input folder.
2. Crop them to the specified size (centered).
3. Save the cropped images to the output folder.
4. Display the first few cropped images as a sample.

Usage as a Python Script

Run this script from the command line, specifying the crop width and height separately. You can also optionally specify the crop start x and y positions, and for videos, the frame interval:

python transform.py --in-directory path/to/input_folder --out-directory path/to/output_folder --crop-width 300 --crop-height 300 --crop-x 50 --crop-y 100 --frame-interval 2.0

Arguments

• --crop-width: The desired width for cropping.
• --crop-height: The desired height for cropping.
• --crop-x: (Optional) X coordinate for the top-left corner of the crop box. If not set, cropping is centered horizontally.
• --crop-y: (Optional) Y coordinate for the top-left corner of the crop box. If not set, cropping is centered vertically.
• --frame-interval: (Optional, videos only) Interval in seconds between frames to extract and crop. Default is 1.0.
• --in-directory: Path to the folder containing the images and/or videos to crop.
• --out-directory: Path to the folder where cropped images will be saved.

Notes

• Images smaller than the specified crop-width or crop-height are skipped, and a message is printed for each skipped image.
• For videos, frames are extracted at the specified interval and saved as PNG images.
• The output directory is created if it doesn’t already exist.
• Supported video formats: .avi, .mp4, .mov, .mkv, .webm
• Supported image formats: .png, .jpg, .jpeg, .bmp

Usage as a Transformation Block in Edge Impulse

This folder has been built to be compatible with Edge Impulse Transfomation block. Simply push this to your project:

edge-impulse-blocks init
edge-impulse-blocks push

For more information about how to configure transformation blocks, refer to Edge Impulse Transformation blocks

Cropped Status Tracking

The crop_image function returns True if the image was cropped and False if it was skipped (i.e., the image is smaller than the crop size). After processing, the script generates an info.labels JSON file in the output directory with metadata for each processed image. The "path" is set to the image filename. "Cropped" in metadata is set to "Yes" if the image was cropped and "No" if it was saved without cropping.

{
    "version": 1,
    "files": [
        {
            "path": "road100.png",
            "category": "split",
            "label": { "type": "unlabeled" },
            "metadata": {
                "Cropped": "Yes"
            }
        },
        {
            "path": "road101.png",
            "category": "split",
            "label": { "type": "unlabeled" },
            "metadata": {
                "Cropped": "No"
            }
        }
    ]
}

Viewing Output

The output folder will contain all the cropped images. In addition, the notebook will display the first few cropped images in a matplotlib plot. Here is an example:

Original:

Cropped:

License

This project is licensed under the MIT License. See the LICENSE file for details.