“"Meh" paper. Shitty name. Am I being pedantic? I think sufficient Machine learning models should refrain from three letter acronyms, there are too many every day! 5-6 letters is better, or it will be unsearchable. No chance I will ever cité this out of fear of confusion with the other tool of the same name. Bonus points for the first team to use emoj…”
GIT: A Generative Image-to-text Transformer for Vision and Language
In this paper, we design and train a Generative Image-to-text Transformer, GIT, to unify vision-language tasks such as image/video captioning and question answering. While generative models provide a consistent network architecture between pre-training and fine-tuning, existing work typically contains complex structures (uni/multi-modal encoder/decoder) and depends on external modules such as object detectors/taggers and optical character recognition (OCR). In GIT, we simplify the architecture as one image encoder and one text decoder under a single language modeling task. We also scale up the pre-training data and the model size to boost the model performance. Without bells and whistles, our GIT establishes new state of the arts on 12 challenging benchmarks with a large margin. For instance, our model surpasses the human performance for the first time on TextCaps (138.2 vs. 125.5 in CIDEr). Furthermore, we present a new scheme of generation-based image classification and scene text recognition, achieving decent performance on standard benchmarks. Codes are released at \url{https://github.com/microsoft/GenerativeImage2Text}.
候选层:至少 2 个不同的人、在 2 个不同月份提到过,还没到主书架的 3 人 3 个月。没有做立场标注,所以只写“被提及”。
HN 书评
没有评论直接链到它;下面是它被提交的讨论串里回复最多的评论(不计入统计)- GIT: A Generative Image-to-Text Transformer for Vision and Language ↗lnyan · 2022-05-30 · 15 分
- Git: A Generative Image-to-Text Transformer for Vision and Language ↗sandkoan · 2023-01-12 · 1 分