An pair of CLI apps written in Python.

• process-video.py - Can split a video into small video files, based on the camera changes found in the movie upload. It can also take an individual movie file, capture it's frames, and save them as .jpg. Uses the PySceneDetect library heavily - https://github.com/dynamite-ready/PySceneDetect.

• evaluate-images.py - Takes a folder of image files, and will determine which of them are NSFW. It's based on ResNet50_NSFW - https://github.com/emiliantolo/pytorch_nsfw_model

You can find more information at - https://raskie.com/post/practical-ai-autodetecting-nsfw

First and foremost, this is a Python repo, so you will need to install it if you don't already have it. The program(s) should build on both Python 2 & 3, but I would recommend version 3 - https://www.python.org/downloads/

If you want to build the standalone CLI executables, you will also need to install PyInstaller separately. You can install it with:

pip install pyinstaller

To run the app in any form, the two following binary files will have to be present in the CWD (Current Working Directory):

• mkvmerge.exe - Is used read and generate MP4 files. It can be found here - https://www.videohelp.com/software/MKVToolNix)

• ResNet50_nsfw_model.pth - The trained weights for the ResNet50NSFW model. The file can be found here - https://github.com/emiliantolo/pytorch_nsfw_model)

A list of all dependent Python modules can be found in requirements.txt. To install the app, you should be able to use the following command:

python -m pip install -r requirements.txt -f https://download.pytorch.org/whl/torch_stable.html

These modules are interdependent, so you need to be wary when they're updated. PyTorch moves particularly fast.

process-video.py / process-video.exe:

usage: process-video.py [-h] [--images] [--skip [SKIP]] [--ipath [IPATH]] file

positional arguments:
  file             A path to the video file you wish to split into smaller videos

optional arguments:
  -h, --help       show this help message and exit
  --images         When set the program will instead capture jpg images from video. Should be used with `--ipath` and `--skip`
  --skip [SKIP]    Number of frames to skip when extracting images
  --ipath [IPATH]  Folder in which to store jpgs when using the `--images` option

evaluate-images.py / evaluate-images.exe:

usage: evaluate-images.py [-h] [--model [MODEL]] folder

positional arguments:
  folder           A folder of images to test for NSFW content. Outputs a single floating point number (0.3+ is probably mucky)

optional arguments:
  -h, --help       show this help message and exit
  --model [MODEL]  `ResNet50_nsfw_model.pth` file location

build-exe.sh or build-exe.bat can be used to create a standalone executable CLI app. You will need to install PyInstaller to run it (see the dependencies section).

• CUDA PyTorch would have been nice, but it's a bloated brick of code
• Take care with the size of the files and folders you want to process
• Is tightly coupled to my Movie Parser GUI project - https://github.com/dynamite-ready/movie-parser