/Few-Shot-Remote-Sensing-Scene-Classification

A collection of surveys and papers, some of which include source code, pertaining to studies related to few-shot learning in remote sensing field.

Survey

  • [JSTARS2021] Research Progress on Few-Shot Learning for Remote Sensing Image Interpretation [Paper]
  • [Remote Sensing 2022] Few-shot object detection in remote sensing image interpretation: Opportunities and challenges [Paper]

Few-shot Image Scene Classification

  • [TGRS2021] DLA-MatchNet for Few-Shot Remote Sensing Image Scene Classification [Paper]

  • [TGRS2021] SPNet: Siamese-Prototype Network for Few-Shot Remote Sensing Image Scene Classification [Paper][Code]

  • [TGRS2022] MKN: Metakernel Networks for Few Shot Remote Sensing Scene Classification [Paper]

  • [TGRS2022] AIFS-DATASET for Few-Shot Aerial Image Scene Classification [Paper]

  • [TGRS2022] SGMNet: Scene Graph Matching Network for Few-Shot Remote Sensing Scene Classification [Paper]

  • [JPRS2022] Task-specific contrastive learning for few-shot remote sensing image scene classification [Paper]

  • [GRSL2022] Learning to cooperate: Decision fusion method for few-shot remote-sensing scene classification [Paper]

  • [Int J Appl Earth Obs 2023] HCPNet: Learning discriminative prototypes for few-shot remote sensing image scene classification [Paper]

  • [TGRS2023] Foreground-background contrastive learning for few-shot remote sensing image scene classification [Paper]

  • [TGRS2023] Multiform ensemble self-supervised learning for few-shot remote sensing scene classification [Paper]

  • [TGRS2023] Multi-pretext-task prototypes guided dynamic contrastive learning network for few-shot remote sensing scene classification [Paper]

Few-shot Object Detection

  • [JPRS2022] Generalized few-shot object detection in remote sensing images [Paper]Code]

Few-shot Domain Generalization

  • [CVPRW2023] APPLeNet: Visual Attention Parameterized Prompt Learning for Few-Shot Remote Sensing Image Generalization using CLIP [Paper][Code]

DataSet

  • NWPU-RESISC45: 31500 images of 45 scene classes (700 per image/class) [Link]
  • AID: 31500 images of 30 scene classes (200~400 per image/class) [Link]
  • UCM: 2100 RS scenes images of 21 classes (100 per image/class) [Link]

DataSplit (Train, Val, and Test)

  • [TGRS2023] Refer to Table I [Paper]