Asirra: a CAPTCHA that exploits interest-aligned manual image categorization

By Jeremy Elson, John R. Douceur, Jon Howell, J. Saul. Conference on Computer and Communications Security, 2007

605 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Information Systems

View Paper: https://www.semanticscholar.org/paper/8c10a7d51d8c33a3daf2c39e16f2e11bf51de55e

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Information Systems

Citations in the corpus

Deep Anomaly Detection Using Geometric Transformations (2018, methodology): “All baseline methods, including OC-SVM variants, which enjoy hindsight information, only achieve performance that is slightly better than random guessing in the CatsVsDogs dataset.”
Anomalous Instance Detection in Deep Learning: A Survey (2020, methodology): “The CatsvsDogs dataset [39], that contains 12,500 images of cats and dogs each, is treated as the OOD data.”
PANDA: Adapting Pretrained Features for Anomaly Detection and Segmentation (2020, methodology): “[3] used a student-teacher based autoencoder approach employing pretrained ImageNet deep features.”
The End is Nigh: Generic Solving of Text-based CAPTCHAs (2014, background): “A significant amount of work has been produced on alternative captchas of this type [36, 20, 17].”
Deep learning multidimensional projections (2019, methodology): “s, ﬂattened to 784-element vectors; Fashion MNIST [XRV17]: 70K observations of 10 types of pieces of clothing, rendered as 28x28-pixel grayscale images, ﬂattened to 784-element vectors; Cats and Dogs [EDHS07]: 25K images of varying sizes divided into two classes (cats, dogs). We used the Inception V”
I am Robot: (Deep) Learning to Break Semantic Image CAPTCHAs (2016, background): “Therefore, such tasks have been considered a promising alternative to text-based captchas, and several schemes have been proposed, in which users are asked to identify or label images [3], [13].”
Improving Shape Deformation in Unsupervised Image-to-Image Translation (2018, methodology): “To demonstrate our network on unaligned data, we evaluate on the Kaggle cat and dog dataset from Microsoft Research [9].”
AUC Maximization in the Era of Big Data and AI: A Survey (2022, background): “Dataset Type Order of Data Size Imbalance Source References for DAM Benchmark Result STL10 Natural Image 10(3) Artiicial [27] [196, 207] CIFAR10 Natural Image 10(4) Artiicial [91] [196, 207] CIFAR100 Natural Image 10(4) Artiicial [91] [196, 207] Cat vs Dog Natural Image 10(4) Artiicial [41] [196, 20”
Structured Pruning Adapters (2022, methodology): “The results of our experiments are presented in Table 1 for model densities of 100% , 30% , and 10% retained weights and visualized for CIFAR-10 in Figure 3 as well as for Oxford Flowers 102 and Cats and Dogs 2 in Appendix C Comparing SPLoRA with ﬁne-pruning, we observed competitive transfer adaptat”
Improving the Robustness of Deep Neural Networks via Adversarial Training with Triplet Loss (2019, neutral): “We conduct experiments over three datasets, i.e., Cats vs. Dogs [Elson et al. , 2007], MNIST [LeCun, 1998] and CI-FAR10 [Krizhevsky and Hinton, 2009].”
Understanding Measures of Uncertainty for Adversarial Example Detection (2018, methodology): “We ﬁnish by demonstrating the effectiveness of dropout on the real-world ASIRRA (Elson et al., 2007) cats and dogs classiﬁcation dataset (Section 4.3).”
Analysis of DAWNBench, a Time-to-Accuracy Machine Learning Performance Benchmark (2018, methodology): “nce. Fine-tuning on new tasks. We additionally ﬁne-tuned a model optimized for time-to-accuracy and a pre-trainedResNet-50providedbyPyTorchonthreeimageclassiﬁcationtasks: Birds-200[47],Kaggle cat/dog [14] (limited to 1000 training and 1000 validation examples), and Leeds butterﬂy [48]. We only train”
Understanding generative AI output with embedding models (2025, unknown): “Despite both samples being composed of real cat images, the vectors from the LSUN (Large-scale Scene UNderstanding) ( 41 ) dataset are markedly distinct from those created from images in the Cats&Dogs ( 42 ) dataset.”