I'm collecting and restoring my family photos and this software is AMAZING, i just cant believe, the color restoration bring all back to life, awesome project , now one to restore from Damage (waves, lines), thanks!
This was quite the excellent walkthrough. I was able to get it working locally by following along. The instructions for the "Installing the Dependencies" section have changed a bit, but the instructions in the repo cover that nicely. Thanks for doing this!
I'm a beginner in programming and I couldn't understand how you opened that window (command prompt) and the path inside it was: " ( deoldifier) D:\RU-vid\............." . can you please explain that...?
This is awesome thanks man for sharing just subscribed keep up the good work! Do you know if it works with videos? Would it separate it frame by frame?
Sir I have a question as you made deep fake face model I have also seen the deep fake audio can we combined both together and make an model please respond sir I am waiting for your reply 😞😅
hey if you remember about your sign language convertor project using jupyter and opencv im getting an error while copy model config to training folder. The syntax of the command is incorrect. please any help?
Appreciate the video, thanks and good work. I could almost follow all of it. I see that it is 11 months old. I work in video and I'm quite interested in setting up a local machine to colorize video. Have you tried this yet with deoldify, or are you interested in trying it? Or are there any other methods that you are familiar with? I am subscribing now, and can't wait to see what you do next!
Hi Nic - fantastic video and great step-by-step instruction - really interesting and really amazing algorithm too, as you say. When I ran the Jupyter notebook I got the following error when running cell #2 ... ModuleNotFoundError: No module named 'fastai' I checked with pip list in my Windows 10 command prompt and fastai is there (1.0.51) - please help as I would dearly love to get this running!
Follow your way step by step, while finally getting the result of "NameError: name 'colorizer' is not defined". What would be the mistake made? How can this be corrected? Thank you.
Can you please show how to train model using image data set in jupyter notebook OR goggle collab (epoch calculation) it will be really helpful for final year project.
Weird, i ran an old photo of my grandmother that had some color in it and it just removed all the color. downloaded black and white photos from the web and it worked as shows.
As i mentioned on your ESRGAN video the other, this is so incredibly fascinating ❤ And super super easy to use, surprisingly fast as well, seems to only take a couple seconds per image at most :) I think it seems to work better with actual greyscale/b&w images, sepia does seem to work okay, but it seems to produce better results when the images are just greyscale, I think. Also, there's probably similar models for doing something like this, but with sound, right? We can deoldify 1920s images now, what about the swing of the roaring 20s? :D (Seriously, don't know what kind of music you're into, but listen to Electro Swing, it'll make you dance, can't help it :D) The following are kind of off-topic thoughts, but you talked a lot about F1 in this video, which reminded me about those thoughts i had a while ago: They are using a ton of wind-tunnel testing with like 60% scale models in F1 nowadays, but also extensive aero simulations probably on their high performance clusters. I wonder if they are also employing AI to sort of predict how a car's aero will perform and I wonder if a well-trained AI model might actually be much less performance/resource hungry than these crazy fluid dynamics sim softwares they're probably using otherwise. I mean, it shouldn't be too difficult to train the AI with data from their simulations to basically "predict" what the sim would output when both are given the same input, right? And I'm sure they must have tons of data already laying around to train such a model with. And I imagine they are probably already using AI to actually design some of their parts (at least if the rules allow them to do so), especially some of the new underfloor designs with those almost wavey patterns to create just the right vortices to me look a little bit like they've been produced by an AI. What do you think, will the new cars succeed in fixing many of the dirty air problems we had? From what I've heard about the tests, it seems pretty promising so far, though they also introduce new problems that come with the venturi effect. Anyways, great video as always, thanks a huge ton! Much appreciated, super awesome :)
I'm sure there's similar for audio restoration as well! Once I'm back to projects I'm hoping to start exploring some of my random ideas again. YESSS, the F1 sim stuff is nuts. Was talking to my boss who's really into it and the data side of things, there is SOOOO much data produced from every hour of running.
Watching your tutorial makes my head hurt, so there is no windows executable yet for this tool? What does this program actually work on? is it Linux only? and is the install process the same for that?
😇 Got it to work: instagram.com/p/ByN-wgioYKv/ 😎 Thanks for covering more about the technical details, Nick! Eventually I'll have the intuition for how GANs work! 😉 Have a great weekend!
@@NicholasRenotte 😳 Thank you,...really: Thank you! Wow! First I thought that was a spam/impersonator account 😂 Oh well 😊 Thanks! Yeah! You know Károly's "Two Minute Papers"-Account?! Think the Colab work book came from there. This is like always that thing: If I look at the latest video about Nvidia's AI-progress, I always wonder how long (or ever?) until this becomes a usable product/thing to utilize. Thanks again!
@@johanneszwilling hahaha nah, I've been trying to get back to replying to comments. Work has been insane so I've been trying to get back to it whenever I can. Yeah I think the bridge from research to production varies depending on the sophistication of the model and it's implementation into open source. E.g. Transformers are way more mainstream now thanks to hugging face and keras!
@@NicholasRenotte Will keep that aspect in mind. Thanks! No worries about replying or not now. I appreciate it! I suppose it’s A LOT of work here and I think there is a healthy balance there where Nick is, with no malice, sometimes just simply more important than that. Thanks again mate!
you know you can take an old black and white movie and break it down into frames and recolor them and then put it back togather haha you can color a black and white 🎥 movie thats wild
hello, its greats, thanks, very good explicated, but i have a question, in my mac, when install openvc or requeriments.txt see this: Encountered error while trying to install package. ╰─> pynvx you can help me? please
