ALIGN: Scaling Up Visual and Vision-Language Representation LearningWith Noisy Text Supervision

Full paper:
arxiv.org/pdf/2102.05918.pdf
Presenter: Nandita Bhaskar
Stanford University, USA
Abstract:
Pre-trained representations are becoming crucial
for many NLP and perception tasks. While representation learning in NLP has transitioned to training on raw text without human annotations, visual and vision-language representations still rely
heavily on curated training datasets that are expensive or require expert knowledge. For vision applications, representations are mostly learned using
datasets with explicit class labels such as ImageNet or OpenImages. For vision-language, popular datasets like Conceptual Captions, MSCOCO,
or CLIP all involve a non-trivial data collection
(and cleaning) process. This costly curation process limits the size of datasets and hence hinders
the scaling of trained models. In this paper, we
leverage a noisy dataset of over one billion image
alt-text pairs, obtained without expensive filtering or post-processing steps in the Conceptual
Captions dataset. A simple dual-encoder architecture learns to align visual and language representations of the image and text pairs using a
contrastive loss. We show that the scale of our
corpus can make up for its noise and leads to
state-of-the-art representations even with such a
simple learning scheme. Our visual representation
achieves strong performance when transferred to
classification tasks such as ImageNet and VTAB.
The aligned visual and language representations
also set new state-of-the-art results on Flickr30K
and MSCOCO benchmarks, even when compared
with more sophisticated cross-attention models.
The representations also enable cross-modality
search with complex text and text + image queries.

[@darkmythos4457]

Thank you, this was a really good summary.

[@stanfordcontrastivesslearn3141]

Glad it was helpful!

[@daeungkim5924]

Thank you, this was really helpful to me to understand the concepts of this paper. I want to share this presentation to the South Korean researchers. Would you mind if I write a blog post about this presentation in Korean and introduce this video(by attaching this video link)?

[@stanfordcontrastivesslearn3141]

Thank you, glad it helps! We would be more than happy that you write a blog post in Korean and link the video. Very nice initiative and good luck with the writing!

[@stanfordcontrastivesslearn3141]

Start at 4:40