Hi Rob, thank you for this tutorial, the demonstration is really good, even as a novice, I can understand it quickly. I also followed your steps to implement it with code, but the resolution of the final generated picture is relatively low (lower than the two original pictures). Is there any good way to improve the clarity of the final picture? Looking forward to your reply, thanks again
Great video, problem is they privatized it. No longer open source, I tried recreating / following along to your video only to find I wasted a chunk of my day. Great content, shame it’s a lost moment in time.
@@robmulla Hi Rob, I am getting error AttributeError: 'INSwapper' object has no attribute 'taskname' when invoking this: app = FaceAnalysis(name='buffalo_l') faces = app.get(image) Any ideas?
Hi Rob, Thanks for the tutorial and want to know if there are any other pre-trained models as the clarity is a bit low. Got a clear picture of the pipe line of face swapping. thanks for your tutorial.
Glad you liked the video. I’m using the solarized dark theme for jupyter lab. I actually have a whole video about my jupyter lab setup you can check out.
Hi Rob. Thanks for the video. In this example, the face is recognized based on a single photo, which can be unflattering. Is it possible to use several photos for more accurate recognition?
Thanks for watching. That's a great question, but I don't believe this package has that capability. I still think the results are impressive given the single photo as input.
I kind of very much disagree--it's probably amazing for beginner *coders* - there were some skipped and unclear parts all over if you don't know how to code and don't know python. The demo title makes it sound super easy "...5 lines Code", but the entire process is far from easy.
Can you suggestions on how to do this in real time video ? Prexisting snapchat filters do a basic swap, not involving the eigen face values and swapping using AI, how can I do this?
Hi Rob. Firts, thanks for the video. I have an issue installing the package: Building wheels for collected packages: insightface Building wheel for insightface (pyproject.toml) ... error error: subprocess-exited-with-error. Did you read something about this? Please, any help would be very well received. Greetings from Peru.
My python version is 3.9 by the way i was able to install packages in my windows machine after install ms c++ 14.0 with visual studio. But in my Ubuntu machine i wasn't cause i didn't find a linux version for VS. Maybe if exist a way to install c++ complements, the problem would be resolved
Hey all, noob question here. Can they see/use/download my photos when i use their package? I do not want to leak my photos and I want to be sure. Thanks
Hi Rob, can you also include the keyboard shortcuts that you are using while showing some of the help topics? it will be very helpful for someone newby like me to follow along Thanks
Great video. I am try to download the inswapper file but i keep getting errors. InvalidProtobuf: [ONNXRuntimeError] : 7 : INVALID_PROTOBUF : Load model from inswapper_128.onnx failed:Protobuf parsing failed. ChatGPT is suggesting that the file is corrupted. please do you have the original file or how do you advice i overcome this?
I am trying this on AWS Ubuntu. However, I am getting this error, even though I have already installed "onnxruntime": sess = C.InferenceSession(session_options, self._model_path, True, self._read_config_from_model) onnxruntime.capi.onnxruntime_pybind11_state.InvalidProtobuf: [ONNXRuntimeError] : 7 : INVALID_PROTOBUF : Load model from ./inswapper_128.onnx failed:Protobuf parsing failed. Can anyone please help me with this?
oh man I can't wait to swap my face onto all of my exgirlfriends photos. its like Im still there, in her life! ㅎㅎㅎㅎㅎ edit: I felt like it was obvious that I'm joking, but it's youtube so I guess I should have lowered my expectations
I'd like to create a face filter. Something like the chadify filter. Would I start with the chad picture and then swap it onto a face? or, would i start wit the face and swap it onto chad's?
Thank u Rob. When i sit down to count my blessings I count "finding robs YT channel" 10 times. Just one request rob. Could you please provide us with the ur YT videos code such as this over github or any ither forum if possible. Would love to take it as a canvas and start experimenting. Thanks again for being my best DATA SCIENCE teacher.
Not representing Insightface or any beneficiary. But be careful about their usage licence - it is tricky. So if anyone gives tutorial video - it is ok. But if monetised - not ok. If you use image for education (what one will do with it - images ! S_x education) - it is ok. If you sell those images - you are not ok. Code is free. Outcome is not...lol... "The code of InsightFace is released under the MIT License. There is no limitation for both academic and commercial usage. The training data containing the annotation (and the models trained with these data) are available for non-commercial research purposes only. Both manual-downloading models from our github repo and auto-downloading models with our python-library follow the above license policy(which is for non-commercial research purposes only)."
@@DrDaab Let me think… An AI that works like a toaster: an on/off lever and a single knob. I shouldn’t need a human to explain me how to make a toast. And if I need to read the manual, it fails the Goddunoff test. :)