Author: Zachary Zeleznick SID: 2324051 Date: 10/06/2015
Functional Directories: Do not remove
| Directories | Subfolders | Description |
|---|---|---|
| data/ | directory from images are loaded | |
| examples/ | a few images that were used in the beginning but later removed from main path | |
| masks/ | a collection of masks used for the mask parameter in main.py | |
| seamData | contains seam indices for precomputation, generated by generateSeamData.py | |
| output/ | directory for images to be saved | |
| basic_images/ | directory containing a few nice images | |
| best_images/ | directory containing my favorite images | |
| bulk_data/ | directory containing past output generated by benchmark.py | |
| logs/ | directory for logs run by generateSeamData.py or by benchmark.py |
Depreciated: No longer in use
| Directories | Description |
|---|---|
| build/ | earlier iterations of code compiled down to C via cython |
| old-code/ | earlier iterations of code for reference |
| notebooks/ | collection of a few tests |
| papers/ | used for research |
-
main.py returns a modified image from user-specified dimensions and optional parameters of saving, masking, and display.
python main.py IMAGE_INDEX = 0; HORIZONTAL_CROP = 0; VERTICAL_CROP = 0; SAVE = False; MASK = False; DISPLAY = True -
benchmark.py generates modified images for all images in the data/ directory with user-specified dimensions and optional parameters of saving, masking, display, and log file output location
python benchmark.py log.txt > debug.txt -
generateSeamData.py generates seam data for for all images in the data/seamData/ directory with either user-specified height or width.
python generateSeamData.py HORIZONTAL_CROP = False -
best.py demonstrates the best sampling of output for the reader
python best.py
-
IO_methods.py handle image and seam loading, reading, and saving.
-
userInput.py handle user input for main.py, benchmark.py and general use
-
utils.py contain useful functions called by main.py and others
-
If the data/ directory have images that have yet to been preprocessed, run
python generateSeamData.py
This will generate 300 seams in the vertical direction and save them in data/seamData/. For horizontal seam generation run
```
python generateSeamData.py horizontal
```
-
Run all the images
python benchmark.py
Current parameters are set to HORIZONTAL_CROP = 100; VERTICAL_CROP = 0; SAVE = True; HAS_MASK = 0; DISPLAY = False
For Display, change the boolean value of DISPLAY to True on line 7. Log file output is defaulted to mainLog.txt, but this can be changed by specifying the file name as the first parameter. More details are displayed to stdout, but can be easily piped to a secondary file (thanks to print override)
- Test a specific image
python main.py 6 0 300 1 1 1
The first argument is the index of the image in data/ to be run. The second and third arguments are the number of pixels to remove from the image. A negative value corresponds to seam insertion, while a positive value is seam removal. The fourth argument is a boolean for optional saving to disk, the fifth is whether to search and load a mask from the data/masks/ directory, which helps to preserve sensitive artifacts such as faces or shadows. The sixth argument is display, which is defaulted to True.
-
See all the Bells & Whistles in Action!
python best.py
Note the difference between masked and unmasked seam removals, the benefit of masks for seam insertion, how quickly pre- proccessed seam removal can be, which pixels are removed and more! See the file for more.
Zachary Zeleznick