top of page


Hateful Memes Contest
-
Developed a multimodal classifier for recognizing hate content (speech, imagery or a combination).
-
The approach involves the development of an intermediate data fusion algorithm, it performs some preprocessing on both modalities of data and then fuses them for input to a downstream classifier block.
-
The data can be viewed by signing up on the challenge URL, please be aware that it contains offensive material, and is not suitable for distribution.
-
The associated entry was ranked in the top 10% of the research phase at the time, and resulted in an opportunity to submit a virtual poster for NeurIPS 2020
-
The associated code sample can be made available on request.
bottom of page
