Resource icon

OBS Python Real-time video censor 0.0.1

Sergey1

New Member
Hi. Thanks for interest in the project.
It depends. If you have an nvidia gpu and installed gpu version of tensorflow, then the model will run on gpu, and image preprocessing will happen on cpu. For image processing it shouldn't be more 5-10% depending on what type of cpu you have. The model itself is "mobilenet" it's supposed to run on mobile phones and doesn't take much ram. Key factor is how fast is you model and image preprocessing, which depends on quiry time set by obs script. By default it's 30 ms which is 30+ fps. If you have powerfull gpu and will lower time to 10-15 it will increase workload on cpu significantly. The idea is you run the script with model when you browse internet and quit when you are done and decide to play game. I won't recommend running model and game at the same time.
If you don't have nvidia gpu, then the model and processing will happen on cpu, which is very expensive and slow.
 

chi11estpanda

New Member
I really like the concept and practical use application. I was wondering, is there any way to have it apply to vehicle license plates or streets too? Seems like it would be a useful feature for IRL streamers with privacy concerns and is hopefully a bit easier to establish an algorithm for prediction thresholds.
 

djbauer

New Member
Hi. Thanks for interest in the project.
It depends. If you have an nvidia gpu and installed gpu version of tensorflow, then the model will run on gpu, and image preprocessing will happen on cpu. For image processing it shouldn't be more 5-10% depending on what type of cpu you have. The model itself is "mobilenet" it's supposed to run on mobile phones and doesn't take much ram. Key factor is how fast is you model and image preprocessing, which depends on quiry time set by obs script. By default it's 30 ms which is 30+ fps. If you have powerfull gpu and will lower time to 10-15 it will increase workload on cpu significantly. The idea is you run the script with model when you browse internet and quit when you are done and decide to play game. I won't recommend running model and game at the same time.
If you don't have nvidia gpu, then the model and processing will happen on cpu, which is very expensive and slow.
Can we just install Tensorflow (GPU) without all the extra steps to install the nvidia stuff? I'm having a hard time understanding the install steps for all the cuda stuff.
 

THE R3Y3S

Member
Hi man, i'm trying to setup this but i can't. I install Tensorflow, setup script and when i press on disable source, it turns black and nothign happen. I can't do nothing and i can't activate it. When i launch classifier.py i can't resize screen so i capture full screen but nothing, on preview i get an image and on cmd i got number response, but on obs still black. How i can fix this? Or what did i do bad?
 
Top