Identifying Creative Harmful Memes via Prompt based Approach
Prerequisites: Google drive account access is necessary for downloading the harm dataset.
It is very simple to use our approach to train & apply a harmful meme detection model: First, use vision-in-text module to generate a caption and a bunch of keywords as mentioned in our paper. Second, use the h-meme model to train a harmful meme detection model, we used harmful-politics dataset as example, which contains the example we used in the following picture.
You can apply to any image-text dataset, thanks to the applicability of this approach. Feel free to check out our original paper: https://dl.acm.org/doi/10.1145/3543507.3587427.
Use citation @inproceedings{10.1145/3543507.3587427, author = {Ji, Junhui and Ren, Wei and Naseem, Usman}, title = {Identifying Creative Harmful Memes via Prompt Based Approach}, year = {2023}, isbn = {9781450394161}, publisher = {Association for Computing Machinery}, address = {New York, NY, USA}, url = {https://doi.org/10.1145/3543507.3587427}, doi = {10.1145/3543507.3587427}, booktitle = {Proceedings of the ACM Web Conference 2023}, pages = {3868–3872}, numpages = {5}, location = {Austin, TX, USA}, series = {WWW '23} }