文章总览 - 164
PRCL:Probabilistic Representation Contrastive Learning for Semi-Supervised Semantic Segmentation
PRCL:Probabilistic Representation Contrastive Learning for Semi-Supervised Semantic Segmentation

发表于IJCV2024,同时是AAAI 2023的oral,将对比学习引入到师生网络,本文提出使用多元高斯分布将像素级表示建模为概率表示(PR)。PR包含一个捕获最可能表示的均值向量和一个表示可靠性的方差向量。PR之间的相似性是通过相互似然评分来衡量的,该评分减少了不确定表示的影响。对于第二个问题,引入了全球分布原型(GDP),以在整个训练过程中聚合全球表示,确保原型位置的一致性。此外,虚拟负片可以从GDP中有效地生成,以补偿零碎的负分布,而不需要内存库。。

11
Region-aware_Contrastive_Learning_for_Semantic_Segmentation
Region-aware_Contrastive_Learning_for_Semantic_Segmentation

12
SegFormer
SegFormer

13
FaceReclaim
FaceReclaim

14
DETR
DETR

15
text1
text1

16
MUN:Image Forgery Localization Based on M3 Encoder and UN Decoder
MUN:Image Forgery Localization Based on M3 Encoder and UN Decoder

发表于AAAI 2025,其使用了Noiseprint++作为低级特征提取器,使用双流结构,其使用池化操作之后的结构作为查询向量,进行融合是最大的创新点。

17
Discriminative fuzzy 𝐾-meansclustering withlocalstructure preservation for high-dimensional data
Discriminative fuzzy 𝐾-meansclustering withlocalstructure preservation for high-dimensional data

18
Omni-IML:Towards Unified Image Manipulation Localization
Omni-IML:Towards Unified Image Manipulation Localization

19
LEGION:Learning to Ground and Explain for Synthetic Image Detection
LEGION:Learning to Ground and Explain for Synthetic Image Detection

20