The split script takes an image as input and cuts the image into fragments using the sliding window approach:
- window size (h, w) in pixels or as a percentage
- x and y offsets
- the result is saved in a directory
- all settings should be included in the names of the resulting images.
The merge script takes a directory containing the segmented images and combines them to reconstruct the original image.
It is necessary to verify that the pixels in the resulting image match the original image exactly.
- What should the sliding window behavior be:
- Add a border when the edges are reached.
- Answer: Shift the window along the axes inside the image to obtain a full fragment that is the same size as the window.
- Windows 11
- Creates folders "input" and "output_split".
- Waits for the input data to be placed in the "input" folder.
- If data exists, the script sets the "input" directory as the current working directory.
- Gets the names of all image files in this directory and stores them in a list.
- Receives input of sliding window parameters. All parameters are by default in pixels, but users can define percentages for window size.
- Checks if the parameters are logically correct; otherwise, brings them to acceptable values. The window size must be lower than the image size, and the offset can't be bigger than the window size.
- Creates the "output_split" folder and makes it the working directory.
- For each image file, creates a sub folder in "output_split" with the image name for its pieces and store the file type in its own name to cover intersections of files with the same name but different types.
- Changes the working directory to the image sub folder.
- For every window, creates a piece of the image file and saves it with parameters in names according to the following pattern:
offs_y_{y}_x_{x}__win_h_{h}_w_{w}__origsize_h_{image_h}_w_{image_w}_{img_file}where:- x, y - offsets
- h, w - window sizes
- image_h, image_w - sizes of source image
- img_file - source image name with file type
- Changes the working directory back to "input" and starts processing the next source image file.
- Creates a folder "output_merged" if it doesn't exist.
- Gets the names of all folders in the "output_split" directory and stores them in a list.
- For every folder, gets the names of all files in the directory and stores them in a list.
- From the first file name, gets the "constant" parameters by template from step 10.
- Using these parameters, creates a canvas with sizes of the source file from "input."
- Loops through pieces of the image and gets their offsets from file names and pastes these pieces into the canvas.
- Saves the image and checks the difference between the source file from "input" and the saved one.
- For the 4th step from above paragraph, it would be great to add a check for files that OpenCV can handle.
- For the 5th step from above paragraph, it would be great to add a switch to choose how to use parameters for the rest of the images.
- Add an option to choose the sliding window type (with or without a border) and implement it in both the split and merge scripts.
- Add directory cleanup for the "output_split" folder to avoid errors when merging:
- On every run of "split.py", make it possible to inspect fragments after merging.
- After a successful merge, since the images are identical, we don't need their fragments anymore.
- As an additional option, add the ability to keep the fragments for annotations, for example.
- Developing/Modifying file processing flowchart: 35 minutes
- Researching relevant information on libraries that work with images: 60 minutes
- Code writing: 40 minutes
- Clarification of questions: 10 minutes
- Making changes: 20 minutes
- Debugging/Testing: 40-60 minutes
- Refactoring: 50 minutes
- Preparing report: 40 minutes
Overall max: 315 minutes = 5.25 hours