bpops / timeslide Goto Github PK

Instead of relying on macOS pinwheel, have window pop up to recognize computation

multiple models for super resolution.

I'm aware that Tkinter shows just a black window on macOS Monterey (possibly Bug Sur). I'm tired of this issue and will be converting over to PyQT instead.

for both macOS and readme.

include tooltips for enhance in GUI

rather than have text entry and button, just button that is "load url from clipboard"

currently, file needs user to explicitly specify ".jpg" in file save dialog.

Currently, resize_image is called on program start, even though there's no image. fix this

Current bootstrap copies the deoldify module from the cloned repo, but better to simply symlink it.

Help > about window for version, dedication info, licenses, etc.

look to resize_image for the correct code.

Hello all,

I am unfortunately not a programmer and was happy to find this program to colorize my old photos.

The installation worked without a problem under Ubuntu 22.04 and the program also starts.

But during the execution it crashes. The following is the message from the console:

 timeslide$ python timeslide.py
NumExpr defaulting to 8 threads.
/home/joern/Github/timeslide/venv/lib/python3.9/site-packages/fastai/data_block.py:442: UserWarning: Your training set is empty. If this is by design, pass `ignore_empty=True` to remove this warning.
  warn("Your training set is empty. If this is by design, pass `ignore_empty=True` to remove this warning.")
/home/joern/Github/timeslide/venv/lib/python3.9/site-packages/fastai/data_block.py:444: UserWarning: Your validation set is empty. If this is by design, use `split_none()`
                 or pass `ignore_empty=True` when labelling to remove this warning.
  warn("""Your validation set is empty. If this is by design, use `split_none()`
Traceback (most recent call last):
  File "/home/joern/Github/timeslide/timeslide.py", line 317, in slideTime
    colorizer = get_image_colorizer(artistic=artistic)
  File "/home/joern/Github/timeslide/deoldify/visualize.py", line 444, in get_image_colorizer
    return get_stable_image_colorizer(root_folder=root_folder, render_factor=render_factor)
  File "/home/joern/Github/timeslide/deoldify/visualize.py", line 453, in get_stable_image_colorizer
    learn = gen_inference_wide(root_folder=root_folder, weights_name=weights_name)
  File "/home/joern/Github/timeslide/deoldify/generators.py", line 19, in gen_inference_wide
    learn.load(weights_name)
  File "/home/joern/Github/timeslide/venv/lib/python3.9/site-packages/fastai/basic_train.py", line 265, in load
    state = torch.load(source, map_location=device)
  File "/home/joern/Github/timeslide/venv/lib/python3.9/site-packages/torch/serialization.py", line 713, in load
    return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
  File "/home/joern/Github/timeslide/venv/lib/python3.9/site-packages/torch/serialization.py", line 920, in _legacy_load
    magic_number = pickle_module.load(f, **pickle_load_args)
_pickle.UnpicklingError: invalid load key, '<'.

Any tips on what I can try?

require tooltips for things that necessitate explanation, e.g., render factor

Need to pretend like nothing happened if no file selected; currently thinks that a file was selected.

If try to save to a file that exists, and select "Replace," python will crash. E.g.,

*** Terminating app due to uncaught exception 'NSInternalInconsistencyException', reason: 'ROSP: sandbox extension failure: nil result or wrong class type ((null)) for finalURL (file:///Users/<username>/Desktop/Untitled, /Users/<username>/Desktop/Untitled)!'

batch mode to do all files with specified parameters from a folder, or multi-file selection.

Pyinstaller now works with Python up to 3.10. Remove that requirement.

Instead of new image simply replacing old image.

Image comparison slider so that user can slide between old and new images. I can't find any examples of this being done in python. Could prove difficult. Might have to use tkinter lines drawn to the canvas, and in realtime crop top image.

currently only have artistic model; should also include stable that might help with zombie limbs

current window is too small; even before we get window scaling implemented, the window should at least be wider to allow for easier analysis of the result.

Canvas should resize with window.

e.g., image-super-resolution

Write bundling script so that this can be standalone macOS app.

Colorize is not the only intended method for this project. Allow for optional enabling/disabling colorizing.

prepare GUI for SR with new frame.

Including all python modules for all aspects of deoldify is unnecessary and creates too large of a bundled app. Remove those that can be.

propose:
step 1: load old photo

• URL or local file
  step 2: parameters
• colorize?
• render factor
• super resolution?
• resolution factor
  step 3: Time slide!
  step 4; save as

Only useful for development, but make window come to front on launch rather than run in background.

URLs always correctly load as the old photo in the canvas, but often the url is rejected when colorized. Do not know reason, but issue is probably somewhere in deoldify's visualize.py?

Provide slider in GUI that allows user to select render_factor (currently hard-coded to 30).

Only show constrained proportion image.