多模态预训练论文
ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks
vision-and-language tasks:
visual question answering,visual commonsense reasoning, referring expressions, and caption-based image
retrieval and a special experiment setting
![](https://img-blog.csdnimg.cn/41508b6a679645d7a538dabb232ebf6d.png)
key technical innovation:
introducing separate streams for vision and language processing that communicate through co-attentional transformer layers.
why two-stream?
![](https://img-blog.csdnimg.cn/bb19535a14c640f6857370c3c791065b.png)
![](https://img-blog.csdnimg.cn/1610e5df35b049d09bd98f6f7d13a585.png)
notes:
Given an image
I
represented as a set of region features v
1
, . . . , v
T
and a text input
w
0
, . . . , w
T
, our model outputs fifinal representations
h
v
0
, . . . , h
v
T
and
h
w
0
, . . . , h
wT
. Notice that
exchange between the two streams is restricted to be between specifific
layers and that the text stream has signifificantly more processing before interacting with visual features
– matching our intuitions that our chosen visual features are already fairly high-level and require
limited context-aggregation compared to words in a sentence.
![](https://img-blog.csdnimg.cn/7ae72a7424a64071b537d40a6eddb092.png)
![](https://img-blog.csdnimg.cn/57f80734205e40fcb4f8d5a5b0f4198e.png)
![](https://img-blog.csdnimg.cn/a50ac51e20ed43b8a84cb12740978072.png)
![](https://img-blog.csdnimg.cn/c4d02e0c5ec84f27a1ca0c7389184751.png)
![](https://img-blog.csdnimg.cn/0b021efc5bd0439c9438fd448e289431.png)
![](https://img-blog.csdnimg.cn/87fdaecf0f3046f4a307940c28406ccd.png)
![](https://img-blog.csdnimg.cn/beb791554296446b8622be69610704b4.png)
The first work is over.
V
ISUAL
BERT: A Simple And Performant Baseline For Vision And Language
two
visually-grounded
language model objectives for pre-training:
(1) part of the text is
masked and the model learns to predict the masked words based on the remaining text and visual
context;
(2) the model is trained to determine whether the provided text matches the image. We
show that such pre-training on image caption data is important for VisualBERT to learn transferable
text and visual representations.
conduct comprehensive experiments on four vision-and-language tasks:VQA VCR NLVR
regionto-phrase grounding
![](https://img-blog.csdnimg.cn/d9f386d4dfe94677b950d69bad4414ab.png)
![](https://img-blog.csdnimg.cn/0bae11b1a4834750bd77c7c9565debc1.png)
![](https://img-blog.csdnimg.cn/a322a1b78fb14948a1c5b5b6e483f9a2.png)
![](https://img-blog.csdnimg.cn/4115f5d6dd6344b6be3c3a828cd7e0e7.png)
![](https://img-blog.csdnimg.cn/ebd147866814432ba77afeb688198ff8.png)
The second work is over.
Unicoder-VL: A Universal Encoder for Vision and Language by Cross-Modal Pre-Training
![](https://img-blog.csdnimg.cn/075e3e2a1a5b49b7b426b7dfbb4c9efd.png)
![](https://img-blog.csdnimg.cn/4669e4515e844db88373c815e0cfd1f4.png)
approach
Pre-training Tasks:MLM MOC VLM
Fine-tune on Downstream Tasks:Image-Text Retrieval.Visual Commonsense Reasoning.and
![](https://img-blog.csdnimg.cn/68f4a11eadda45859d095a1063ba5471.png)
![](https://img-blog.csdnimg.cn/451c08026c71425484027369215da46e.png)
The third word is over.
LXMERT: Learning Cross-Modality Encoder Representations from Transformers
It consists of three Transformer
: encoders: an object
relationship encoder, a language encoder, and across-modality encoder.
pre-train our model with fifive diverse representative tasks:
(1) masked cross
modality language modeling
(2) masked object
prediction via RoI-feature regression
(3) masked
object prediction via detected-label classifification,
(4) cross-modality matching
(5) image question answering.
![](https://img-blog.csdnimg.cn/8b1666f2fa80420eaee8be1a7f2338f6.png)
![](https://img-blog.csdnimg.cn/5d73c7095f694cf18c4720fa9e9022bd.png)
![](https://img-blog.csdnimg.cn/bdea3d4c27a343c49fbbec8bfe4d4d74.png)
![](https://img-blog.csdnimg.cn/d8abf36ebc8d4824a8f905b1406b29e3.png)
over