Task
對某一類別c1，選擇已有標註的樣本作爲support，待測樣本爲 query，model 以 support 爲參考去分割 query。
C-way K-shot：C爲類別數；分割每個 query 使用 K 個 support images
Solution
提取support 和 query 的 feature，將兩者的 feature 進行比對。訓練時藉助其他類別的標註數據去訓練model。
數據集
pascal-5i
每個sub-dataset中，trainset來自pascal trainset，testset來自pascal valset；trainset 和 testset 的類別無交集
One-Shot Learning for Semantic Segmentation (2017.09)
dataset： pascal-5i
train時，support 和 query 均從Dtrain中抽樣，class相同
test時，sample N = 1000 examples 
Metric：mIoU
結構：conditioning branch 輸出參數 θ 用於segmentation branch 輸出feature 的分類
for k-shot：對每個support單獨得到的mask做logical OR
SG-One: Similarity Guidance Network for One-Shot Semantic Segmentation （2018.11）
query image 送入 Guidance Branch 和 Segmentation Branch，通過concate 來使 Guidance Branch 輸出指導Segmentation Branch。
CosineSimilarty map ：as guidance to teach the segmentation branch to discover the desired object regions.
for k-shot：
1、
Y is the predicted semantic label
2、use the averaged vector to guide the segmentation process
CANet: Class-Agnostic Segmentation Networks with Iterative Refinement and Attentive Few-Shot Learning （2019.03）
Dense Comparison Module：
Feature Extractor：ResNet-50，使用中間 block 的 feature（middle-level features that may
constitute object parts shared by unseen classes）
use dilated convolutions in layers after block2
Dense Comparison：concatenated feature maps go through another convolutional block with 256 3 × 3 convolutional filters for comparison.
Iterative Optimization Module
循環迭代，alternatively use predicted masks in the last epoch and empty masks as the input to IOM（dropout）
for k-shot：The result from the attention branch serves as the weight λ, λ 通過 softmax 作爲 sum 權重
使用弱標籤
PANet: Few-Shot Image Semantic Segmentation with Prototype Alignment (2019.08 ICCV)
與檢測的 PANet 無關，重名了
support 和 query image 公用 feature extractor，對 support feature 做 mask GAP 得到 prototypes。
query feature 與 prototypes 計算 cosine distance
Prototype alignment regularization (PAR)：將support 和 query 互換，taking the query and the predicted mask as the new support to learn to segment the support images.
for k-shot：所有 support 一起做 Mask GAP
