Tiny ImageNet Visual Recognition Challenge

By Ya Le, Xuan S. Yang. 2015

Subfields: Computer Vision and Pattern Recognition, Evaluation & Benchmarks

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In Evaluation & Benchmarks

Citations in the corpus

SGDR: Stochastic Gradient Descent with Warm Restarts (2016, neutral): “In contrast to earlier attempts (Pouransari & Ghili, 2015), our downsampled ImageNet contains exactly the same images from 1000 classes as the original ImageNet but resized with box downsampling to 32 × 32 pixels.”
Greedy AutoAugment (2019, methodology): “In the experiments, we also use four real datasets, 1- Tiny ImageNet[35] includes 120000 natural images in 200 classes with each class having a training set of 500 images a test set of 50 images along with 50 validation images.”
Learnability and Algorithm for Continual Learning (2023, neutral): “Tiny-ImageNet (Le & Yang, 2015).”
Task-Aware Information Routing from Common Representation Space in Lifelong Learning (2023, methodology): “We obtain Seq-CIFAR10, Seq-CIFAR100 and Seq-TinyImageNet by splitting CIFAR10 (Krizhevsky et al., 2009), CIFAR100 (Krizhevsky et al., 2009) and TinyImageNet (Le & Yang, 2015) into 5, 5 and 10 partitions of 2, 20 and 20 classes per task, respectively.”
FeTrIL: Feature Translation for Exemplar-Free Class-Incremental Learning (2022, methodology): “We use four public datasets: (1) CIFAR-100 [16] - 100 classes, 32x32 pixels images, 500 and 100 images/class for training and test; (2) TinyImageNet [18] - 200 leaf clases from ImageNet, 64x64 pixels images, 500 and 50 for training and test; (3) ImageNet-Subset - 100 classes subset of ImageNet…”
Task-Free Continual Learning via Online Discrepancy Distance Learning (2022, neutral): “5 In the following, we also evaluate the performance of the proposed approach on the large-scale dataset (MINIImageNet [21]), and Permuted MNIST [9].”
Class-incremental Novel Class Discovery (2022, methodology): “We have used three data sets to conduct experiments for class-iNCD: CIFAR-10 [23], CIFAR-100 [23] and Tiny-ImageNet [26].”
Federated Class-Incremental Learning (2022, methodology): “We use three datasets: CIFAR-100 [13, 21], ImageNet-Subset [7], and TinyImageNet [35] in our experiments.”
Self-Sustaining Representation Expansion for Non-Exemplar Class-Incremental Learning (2022, methodology): “As shown in Table 3, compared to the SOTA of non-exemplar methods (E=0), our method achieves average improvement of 3, 3 and 6 points on CIFAR-100, TinyImageNet and ImageNetSubset, respectively.”
Learning Fast, Learning Slow: A General Continual Learning Method based on Complementary Learning System (2022, methodology): “Following Buzzega et al. (2020a) (Pouransari & Ghili, 2015) ( S-Tiny-ImageNet ) which are split into 5, 5, and 10 tasks each including 2, 2, and 20 classes respectively.”
GCR: Gradient Coreset based Replay Buffer Selection for Continual Learning (2021, positive): “Epochs are increased to 100 for Tiny-ImageNet, which is commonly done for harder datasets.”
Recent Advances of Continual Learning in Computer Vision: An Overview (2021, neutral): “We also give a brief discussion on some potential future research directions.”
Always Be Dreaming: A New Approach for Data-Free Class-Incremental Learning (2021, methodology): “Additional experiments on the challenging TinyImageNet dataset [30], which demonstrate the scalability of our method, are available in the Appendix.”
Split-and-Bridge: Adaptable Class Incremental Learning within a Single Neural Network (2021, methodology): “As a labeled test set of Tiny-ImageNet is not available, we use its validation set as a test set.”
Layerwise Optimization by Gradient Decomposition for Continual Learning (2021, neutral): “The split tiny ImageNet is derived from tiny ImageNet [24].”
Memory-Efficient Semi-Supervised Continual Learning: The World is its Own Replay Buffer (2021, methodology): “We report results for the Tiny-ImageNet dataset [24], which contains 200 classes of 64x64 resolution images with 500 training images per class, in Table IIIb.”