GIT: A Generative Image-to-text Transformer for Vision and Language
This work addresses the need for a simpler and more effective model for vision-language tasks like captioning and question answering, though it is incremental in its architectural simplification and scaling.
The authors tackled the problem of unifying vision-language tasks by proposing GIT, a simplified generative image-to-text transformer, which achieved new state-of-the-art results on 12 benchmarks, including surpassing human performance on TextCaps with a CIDEr score of 138.2 vs. 125.5.
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}.