Skip to content
This repository was archived by the owner on Dec 25, 2022. It is now read-only.

Latest commit

 

History

History
99 lines (65 loc) · 2.99 KB

README-CUDA.md

File metadata and controls

99 lines (65 loc) · 2.99 KB

Information regarding getting RecFilter v2 to use an NVIDIA GPU.

Credits to @Datahell for working all this out.

Requirement:

Python 3.7.9 -> 3.9.6 - Has been tested on 3.7.9 and 3.9.5

A NVIDIA GPU with CUDA.

Installation:

RecFilter2 and NudeNet:

Clone/download the repo, extract to a directory, then open a console/terminal within that directory.

Install the dependencies by entering:

pip install -r requirements.txt

or

python -m pip install -r requirements.txt

If you are running Python 3.8/3.9 on Windows 10 then open a console and enter the following commands:

python -m pip uninstall protobuf
python -m pip install protobuf

This will ensure you are using the latest version of protobuf which fixes a bug with running slow on Windows 10.

CUDA Drivers:

Make sure you have the latest nVida GFX drivers for your card installed.

Install the latest version of the CUDA Toolkit available from NVIDIA.

Install the latest version of the cuDNN libraries, this requires registering for a NVIDIA Developer account, (free).

Go to cuDNN download, you will need to login, agree to the license, then select your download.

NOTE: It gives a download for Windows (x86) but it's actually Windows (x64) libraries.

Install the cuDNN as per the instructions: cuDNN Installation Guide

Reboot your computer.

Additional Python Modules:

Open a console/terminal:

We need to replace onnxruntime with onnxruntime-gpu to utilise CUDA:

python -m pip uninstall -y onnxruntime
python -m pip install onnxruntime-gpu==1.8.1

We need to install the TensorFlow modules to convert the model:

python -m pip install tensorflow-gpu
python -m pip install tf2onnx

Conversion of the onnx model

Download the NudeNet detector checkpoint archive, NudeNet Detector Checkpoint, and extract to a directory.

Find the file detector_v2_default_checkpoint.onnx on your system, for Windows this will be:

C:\Users\<username>\.NudeNet\detector_v2_default_checkpoint.onnx.

Rename it to detector_v2_default_checkpoint.onnx.backup.

Windows:

ren detector_v2_default_checkpoint.onnx detector_v2_default_checkpoint.onnx.backup

Linux:

mv detector_v2_default_checkpoint.onnx detector_v2_default_checkpoint.onnx.backup

Convert the checkpoint file (Windows):

python -m tf2onnx.convert --saved-model <path to extracted archive>\detector_v2_default_checkpoint_tf --opset 11 --output <path to original checkpoint file>\detector_v2_default_checkpoint.onnx

This should only take a couple of minutes, (depending on your hardware), and will probably generate a few warnings but it will still succeed.

At this point the NudeNet API should use CUDA to perform image analysis.