Official implementation is released by the authors.
This is an unofficial implementation of the paper, Zero-shot Visual Commonsense Immorality Prediction [Jeong+, BMVC2022].
Note that the paper might contain images and descriptions of an offensive nature and that this repository uses data described in the paper.
- Python 3.8+
- PyTorch (tested with 1.12.1)
conda create -n zsvcip python=3.8
conda activate zsvcip
conda install pytorch==1.12.1 torchvision==0.13.1 torchaudio==0.12.1 cudatoolkit=11.3 -c pytorch
pip install -r requirements.txt
pip install -e .
This repository provides a script to prepare the ETHICS dataset. See datasets/README.md for more details.
python datasets/prepare_ethics.py
python tools/train.py
To change configuration from the command line, type "--" followed by a space-separated list of keys and values.
python tools/train.py \
-- \
input.batch_size 16 \
model.clip_model openai/clip-vit-base-patch16
python tools/evaluate.py \
-- \
resume outputs/latest.pth
python tools/inference.py \
-i 'hello world' \
-m text \
-- \
resume outputs/latest.pth
For zero-shot prediction, this repository provides a code to download images from Bing by specifying keywords.
python tools/image_crawler.py \
--root_dir cat \
--keyword 'cat' \
--license 'creativecommons' \
-n 1
To input an image into the network, it is necessary to change the mode and the network architecture as follows:
python tools/inference.py \
-i cat/000001.jpg \
-m image \
-- \
resume outputs/latest.pth \
model.arch image
@inproceedings{Jeong_2022_BMVC,
author = {Yujin Jeong and Seongbeom Park and Suhong Moon and Jinkyu Kim},
title = {Zero-shot Visual Commonsense Immorality Prediction},
booktitle = {33rd British Machine Vision Conference 2022, {BMVC} 2022, London, UK, November 21-24, 2022},
publisher = {{BMVA} Press},
year = {2022},
url = {https://bmvc2022.mpi-inf.mpg.de/0320.pdf}
}
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent