Hi everyone, I really appreciate the positive feedback! I figured that it would be useful for me to let people know that my dissertation writing is currently in full swing, so I might not get back to creating additional tutorials for a while, but rest assured that I will be coming back to it later this year. Please feel free to continue leaving feedback and suggestions on topics to cover, and I'll try to get to them once I have the time to do it!
The prediction structure you coded the prediction value as an int, is it fixed structure for all models ? For instance, I am using yolov4 ONNX model to predict objects within my 3D scene, will the prediction structure be the same as yours ? or am I gonna need to code different structure in order to accommodate yolov4's output. I am new to machine learning so I don't know whether outputs differ based on models.
For anyone else struggling a bit to install the barracuda package, from their GitHub: Note: in Unity 2020.3 and later Barracuda is not visible in the Package Manager UI. It can be enabled by manually editing your project's Packages/manifest.json and adding a com.unity.barracuda entry there. Starting in 2021.1 you can add the package directly through the Package Manager interface by selecting + and then Add package by name...
I'm glad you liked it! I'd like to do more tutorials, but in all honesty, since this video went up, I defended and graduated. Right now my focus is on my new job and getting the remainder of my dissertation work published with peer-reviewed journals/conferences, so it will realistically be a while before I can get back to making more. It was a lot of fun to make this video and I hope to make more soon - I do read every comment, so I will be using them as input for what I make next!
Hi Alec, thanks for putting this together. Do I need a GPU to run this, I mean a dedicated GPU on my laptop/desktop as opposed to an integrated graphics processor like Intel HD graphics. In the video you mention Barracuda puts the neural network stuff onto a texture and passes it on to the GPU to effectively run as a compute shader, and I noticed you do Device.GPU in your code, I just wonder if this can run on an IGP or needs a GPU. Thanks and all the best.
Hi Darío, I don't have a machine with an integrated GPU to test on, but from my understanding, Barracuda ought to be running through the same pipeline as regular graphics shaders and compute shaders in Unity. So while it might not run as quickly as with a dedicated graphics card, you shouldn't need to modify anything for this to run.
Hi, great video, just one question.. can we create an android build/apk for this project and run it on actual device, i tried using onxx models in some other project but wasn't able to generate a build for android. Please let me know!
HI thank you for the great content I need to ask you something iam doing a 3d hand controlles by machine learning for graduation can i please know what i need? thanks in advance
Hi everyone, I really appreciate the positive feedback! I figured that it would be useful for me to let people know that my dissertation writing is currently in full swing, so I might not get back to creating additional tutorials for a while, but rest assured that I will be coming back to it later this year. Please feel free to continue leaving feedback and suggestions on topics to cover, and I'll try to get to them once I have the time to do it!
Hi! Thank you for sharing your knowledge.
Can I use a model I trained myself for pictures?
This is the course I was looking for. Thank you so much.
This video is really helpful !! amazing tutorial for a both Unity&Barracuda rookie. Thanks for all your work Alec!
Really Really Great Video Thank you for Uploading this Video
Gran video amigo,me fue de mucha ayuda.Gracias
We need more of these lessons. Can you teach us about hand movement recognition?
The prediction structure you coded the prediction value as an int, is it fixed structure for all models ? For instance, I am using yolov4 ONNX model to predict objects within my 3D scene, will the prediction structure be the same as yours ? or am I gonna need to code different structure in order to accommodate yolov4's output. I am new to machine learning so I don't know whether outputs differ based on models.
For anyone else struggling a bit to install the barracuda package, from their GitHub: Note: in Unity 2020.3 and later Barracuda is not visible in the Package Manager UI. It can be enabled by manually editing your project's Packages/manifest.json and adding a com.unity.barracuda entry there. Starting in 2021.1 you can add the package directly through the Package Manager interface by selecting + and then Add package by name...
Wow great video, thanks a lot for sharing!! Will you do more tutorials on barracuda or something? Thanks!!
I'm glad you liked it! I'd like to do more tutorials, but in all honesty, since this video went up, I defended and graduated. Right now my focus is on my new job and getting the remainder of my dissertation work published with peer-reviewed journals/conferences, so it will realistically be a while before I can get back to making more. It was a lot of fun to make this video and I hope to make more soon - I do read every comment, so I will be using them as input for what I make next!
Thanks man =)@@tapiralec
Hi Alec, thanks for putting this together. Do I need a GPU to run this, I mean a dedicated GPU on my laptop/desktop as opposed to an integrated graphics processor like Intel HD graphics. In the video you mention Barracuda puts the neural network stuff onto a texture and passes it on to the GPU to effectively run as a compute shader, and I noticed you do Device.GPU in your code, I just wonder if this can run on an IGP or needs a GPU. Thanks and all the best.
Hi Darío, I don't have a machine with an integrated GPU to test on, but from my understanding, Barracuda ought to be running through the same pipeline as regular graphics shaders and compute shaders in Unity. So while it might not run as quickly as with a dedicated graphics card, you shouldn't need to modify anything for this to run.
Hi, great video, just one question.. can we create an android build/apk for this project and run it on actual device, i tried using onxx models in some other project but wasn't able to generate a build for android. Please let me know!
goat fr
I need this but with a letter-recognition model, for "q", "p", "d", "b". Please, I need your help to find one.
This is so cool!!!
it is really coool! thanks for sharing
HI thank you for the great content
I need to ask you something iam doing a 3d hand controlles by machine learning for graduation can i please know what i need?
thanks in advance
when will we get more tutorials?
phd student in vr and machine learning? same here! good luck
Thank you so much.
How to do prediction for Alphabets?
so cool!thank you