mlsanityData to 7 Oct 2026

Learning to Obstruct Few-Shot Image Classification over Restricted Classes

arXiv 2409.19210 ↗First posted 28 Sep 2024cs.CVcited by 5

Not on the map. Why?

Abstract (arXiv)

Advancements in open-source pre-trained backbones make it relatively easy to fine-tune a model for new tasks. However, this lowered entry barrier poses potential risks, e.g., bad actors developing models for harmful applications. A question arises: Is possible to develop a pre-trained model that is difficult to fine-tune for certain downstream tasks? To begin studying this, we focus on few-shot classification (FSC). Specifically, we investigate methods to make FSC more challenging for a set of restricted classes while maintaining the performance of other classes. We propose to meta-learn over the pre-trained backbone in a manner that renders it a ''poor initialization''. Our proposed Learning to Obstruct (LTO) algorithm successfully obstructs four FSC methods across three datasets, including ImageNet and CIFAR100 for image classification, as well as CelebA for attribute classification.

Key facts. Cited by 5 papers in the map (data through 7 Oct 2026). Builds on 48 of its 64 references that are in the map. First posted on arXiv 28 Sep 2024.

Builds on

48 of 64 references.

References · 48

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  3. 2017Mask R-CNNcited by 7,259
  4. 2016Densely Connected Convolutional Networkscited by 6,691
  5. 2017Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networkscited by 4,861
  6. 2014Fully Convolutional Networks for Semantic Segmentationcited by 4,845
  7. 2018MobileNetV2: Inverted Residuals and Linear Bottleneckscited by 4,773
  8. 2014Deep Learning Face Attributes in the Wildcited by 3,679
  9. 2017Prototypical Networks for Few-shot Learningcited by 3,001
  10. 2021Learning to Prompt for Vision-Language Modelscited by 2,269
  11. 2016Matching Networks for One Shot Learningcited by 2,246
  12. 2017Understanding Black-box Predictions via Influence Functionscited by 1,714
36 more
  1. 2016YOLO9000: Better, Faster, Strongercited by 1,429
  2. 2022Conditional Prompt Learning for Vision-Language Modelscited by 1,427
  3. 2017Learning to Compare: Relation Network for Few-Shot Learningcited by 1,139
  4. 2021CLIP-Adapter: Better Vision-Language Models with Feature Adapterscited by 1,021
  5. 2019A Closer Look at Few-shot Classificationcited by 679
  6. 2018Do Better ImageNet Models Transfer Better?cited by 560
  7. 2019Meta-Learning with Implicit Gradientscited by 523
  8. 2018Poison Frogs! Targeted Clean-Label Poisoning Attacks on Neural Networkscited by 500
  9. 2019Meta-Learning with Differentiable Convex Optimizationcited by 461
  10. 2018TADAM: Task dependent adaptive metric for improved few-shot learningcited by 408
  11. 2021Supervision Exists Everywhere: A Data Efficient Contrastive Language-Image Pre-training Paradigmcited by 405
  12. 2022Tip-Adapter: Training-free Adaption of CLIP for Few-shot Classificationcited by 355
  13. 2021VL-Adapter: Parameter-Efficient Transfer Learning for Vision-and-Language Taskscited by 319
  14. 2019A backdoor attack against LSTM-based text classification systemscited by 214
  15. 2021Fast Yet Effective Machine Unlearningcited by 209
  16. 2016What makes ImageNet good for transfer learning?cited by 192
  17. 2018Low-Shot Learning from Imaginary Datacited by 181
  18. 2020Meta-Baseline: Exploring Simple Meta-Learning for Few-Shot Learningcited by 179
  19. 2023Fast Machine Unlearning Without Retraining Through Selective Synaptic Dampeningcited by 177
  20. 2020Mixed-Privacy Forgetting in Deep Networkscited by 158
  21. 2018Task-Agnostic Meta-Learning for Few-shot Learningcited by 151
  22. 2022Pushing the Limits of Simple Pipelines for Few-Shot Learning: External Data and Fine-Tuning Make a Differencecited by 123
  23. 2017Few-Shot Image Recognition by Predicting Parameters from Activationscited by 122
  24. 2020Few-Shot Classification with Feature Map Reconstruction Networkscited by 116
  25. 2022CLIP Models are Few-shot Learners: Empirical Studies on VQA and Visual Entailmentcited by 113
  26. 2020Just How Toxic is Data Poisoning? A Unified Benchmark for Backdoor and Data Poisoning Attackscited by 109
  27. 2019TapNet: Neural Network Augmented with Task-Adaptive Projection for Few-Shot Learningcited by 90
  28. 2021Are Gender-Neutral Queries Really Gender-Neutral? Mitigating Gender Bias in Image Searchcited by 90
  29. 2018Delta-encoder: an effective sample synthesis method for few-shot object recognitioncited by 89
  30. 2018Is feature selection secure against training data poisoning?cited by 88
  31. 2021Relational Embedding for Few-Shot Classificationcited by 74
  32. 2019Image Deformation Meta-Networks for One-Shot Learningcited by 56
  33. 2022Few-Shot Classification with Contrastive Learningcited by 31
  34. 2020Few-Shot One-Class Classification via Meta-Learningcited by 17
  35. 2023IMMA: Immunizing text-to-image Models against Malicious Adaptationcited by 17
  36. 2022Adversarial Attacks are a Surprisingly Strong Baseline for Poisoning Few-Shot Meta-Learnerscited by 3

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 1; 2025: 3; 2026: 1.

5 papers.

  1. 2025Model Immunization from a Condition Number Perspectivecited by 8
  2. 2025Knowledge Distillation Detection for Open-weights Modelscited by 4
  3. 2026Designing to Forget: Deep Semi-parametric Models for Unlearningcited by 1
  4. 2024ResidualDroppath: Enhancing Feature Reuse over Residual Connectionscited by 0
  5. 2025Color as the Impetus: Transforming Few-Shot Learnercited by 0