Spaces:
Runtime error
Apply for community grant: Personal project
Hello HF Team and Community,
I'm applying for a GPU community grant as part of this project on the colorization of black-and-white images and videos.
I've currently implemented 2 models based on Convolutional Neural Networks, one from ECCV16 and another from SIGGRAPH17. This space allows the user to either upload a batch of images, a video file, or a youtube link and returns a colorized version of it.
At the moment all these features are implemented and working correctly. However, as soon as the user wants to colorize a video the CPU Basic greatly limits the user experience, since to colorize a video, I have to colorize every frame of the video, and that takes a lot of time.
See the example below of the colorization of a one-minute black & white video by Charlie Chaplin.
So, to overcome this problem and to allow users to choose other, more complex, and qualitative GAN-based models that I plan to implement, access to a GPU is essential for this application.
Thank you for taking the time to read my request and I hope to hear from you soon!
Clément.
Must say, this is truly amazing and it definitely gets the job done. However, did encounter a small problem when dealing with long videos. You see, having a collection of black and white movies, would love to experiment with them by adding a touch of dreamy colors using AI colorization. The issue is that every time the process begins, the estimated time keeps fluctuating between 3 days and 6 days, and so on until it stops working. Is there any chance to obtain or download the software to use it offline?
Wreckless4415 - https://archive.org/details/@jerry_r910
Cyborg Sample:
Download - https://fastupload.io/en/AiFPy7lmRr80A5m/fileHello
@Wreckless4415
I will update the installation instructions of the original repo ASAP : https://github.com/Wazzabeee/image-video-colorization
In the meantime you can follow this :
If you want to process longer videos and you're limited by the Hugging Face space memory's limits, you can run this app locally.
ffmpeg.exe is needed to run this app, you can install it using brew install ffmpeg and update the IMAGEIO_FFMPEG_EXE environment variable accordingly.
git clone https://github.com/Wazzabeee/image-video-colorization
cd image-video-colorization
pip install -r requirements.txt
streamlit run 01_📼_Upload_Video_File.py
That way you can run the app locally :)