The dataset used in the paper is CIFAR-10, CIFAR-100 and Tiny-Imagenet datasets.

The CIFAR-10 dataset consists of 60000 32x32 colour images in 10 classes, with 6000 images per class. There are 50000 training images and 10000 test images.

To use this dataset:

import tensorflow_datasets as tfds

ds = tfds.load('cifar10', split='train')
for ex in ds.take(4):
 print(ex)

See the guide for more informations on tensorflow_datasets.

https://storage.googleapis.com/tfds-data/visualization/fig/cifar10-3.0.2.png" alt="Visualization" width="500px">

CINIC-10 is a dataset for image classification. It has a total of 270,000 images, 4.5 times that of CIFAR-10. It is constructed from two different sources: ImageNet and CIFAR-10. Specifically, it was compiled as a bridge between CIFAR-10 and ImageNet. It is split into three equal subsets - train, validation, and test - each of which contain 90,000 images.

Towards Realistic Out-of-Distribution Detection: A Novel Evaluation Framework for Improving Generalization in OOD Detection:

This paper presents a novel evaluation framework for Out-of-Distribution (OOD) detection that aims to assess the performance of machine learning models in more realistic settings. We observed that the real-world requirements for testing OOD detection methods are not satisfied by the current testing protocols. They usually encourage methods to have a strong bias towards a low level of diversity in normal data. To address this limitation, we propose new OOD test datasets (CIFAR-10-R, CIFAR-100-R, and ImageNet-30-R) that can allow researchers to benchmark OOD detection performance under realistic distribution shifts. Additionally, we introduce a Generalizability Score (GS) to measure the generalization ability of a model during OOD detection. Our experiments demonstrate that improving the performance on existing benchmark datasets does not necessarily improve the usability of OOD detection models in real-world scenarios. While leveraging deep pre-trained features has been identified as a promising avenue for OOD detection research, our experiments show that state-of-the-art pre-trained models tested on our proposed datasets suffer a significant drop in performance. To address this issue, we propose a post-processing stage for adapting pre-trained features under these distribution shifts before calculating the OOD scores, which significantly enhances the performance of state-of-the-art pre-trained models on our benchmarks.

The dataset used in the paper is CIFAR-10, Imagenette, and ImageNet.

CINIC-10 is an augmented extension of CIFAR-10. It contains the images from CIFAR-10 (60,000 images, 32x32 RGB pixels) and a selection of ImageNet database images (210,000 images downsampled to 32x32). It was compiled as a 'bridge' between CIFAR-10 and ImageNet, for benchmarking machine learning applications. It is split into three equal subsets - train, validation, and test - each of which contain 90,000 images.

This repository contains the data from the paper, "Benchmark Generation Framework with Customizable Distortions for Image Classifier Robustness."

Relevant URLs:

https://hewlettpackard.github.io/trust-ml/

https://github.com/HewlettPackard/trust-ml/

Abstract:

We present a novel framework for generating adversarial benchmarks to evaluate the robustness of image classification models. The RLAB framework allows users to customize the types of distortions to be optimally applied to images, which helps address the specific distortions relevant to their deployment. The benchmark can generate datasets at various distortion levels to assess the robustness of different image classifiers. Our results show that the adversarial samples generated by our framework with any of the image classification models, like ResNet-50, Inception-V3, and VGG-16, are effective and transferable to other models causing them to fail. These failures happen even when these models are adversarially retrained using state-of-the-art techniques, demonstrating the generalizability of our adversarial samples. Our framework also allows the creation of adversarial samples for non-ground truth classes at different levels of intensity, enabling tunable benchmarks for the evaluation of false positives. We achieve competitive performance in terms of net $L_2$ distortion compared to state-of-the-art benchmark techniques on CIFAR-10 and ImageNet; however, we demonstrate our framework achieves such results with simple distortions like Gaussian noise without introducing unnatural artifacts or color bleeds. This is made possible by a model-based reinforcement learning (RL) agent and a technique that reduces a deep tree search of the image for model sensitivity to perturbations, to a one-level analysis and action. The flexibility of choosing distortions and setting classification probability thresholds for multiple classes makes our framework suitable for algorithmic audits.

CIFAR-100

The CIFAR-10 and CIFAR-100 dataset contains labeled subsets of the 80 million tiny images dataset. They were collected by Alex Krizhevsky, Vinod Nair, and Geoffrey Hinton. * More info on CIFAR-100: https://www.cs.toronto.edu/~kriz/cifar.html * TensorFlow listing of the dataset: https://www.tensorflow.org/datasets/catalog/cifar100 * GitHub repo for converting CIFAR-100 tarball files to png format: https://github.com/knjcode/cifar2png

All images were sized 32x32 in the original dataset

The CIFAR-10 dataset consists of 60,000 32x32 colour images in 10 classes, with 6,000 images per class. There are 50,000 training images and 10,000 test images [in the original dataset].

This dataset is just like the CIFAR-10, except it has 100 classes containing 600 images each. There are 500 training images and 100 testing images per class. The 100 classes in the CIFAR-100 are grouped into 20 superclasses. Each image comes with a "fine" label (the class to which it belongs) and a "coarse" label (the superclass to which it belongs). However, this project does not contain the superclasses. * Superclasses version: https://universe.roboflow.com/popular-benchmarks/cifar100-with-superclasses/

More background on the dataset: https://i.imgur.com/5w8A0Vm.png" alt="CIFAR-100 Dataset Classes and Superclassees">

Version 1 (original-images_Original-CIFAR100-Splits):

• Original images, with the original splits for CIFAR-100: train (83.33% of images - 50,000 images) set and test (16.67% of images - 10,000 images) set only.
• This version was not trained

Version 2 (original-images_trainSetSplitBy80_20):

• Original, raw images, with the train set split to provide 80% of its images to the training set (approximately 40,000 images) and 20% of its images to the validation set (approximately 10,000 images)
• Trained from Roboflow Classification Model's ImageNet training checkpoint
• https://blog.roboflow.com/train-test-split/ https://i.imgur.com/kSPeKGn.png" alt="Train/Valid/Test Split Rebalancing">

Citation:

@TECHREPORT{Krizhevsky09learningmultiple,
  author = {Alex Krizhevsky},
  title = {Learning multiple layers of features from tiny images},
  institution = {},
  year = {2009}
}

ImageNet-P consists of noise, blur, weather, and digital distortions. The dataset has validation perturbations; has difficulty levels; has CIFAR-10, Tiny ImageNet, ImageNet 64 × 64, standard, and Inception-sized editions; and has been designed for benchmarking not training networks. ImageNet-P departs from ImageNet-C by having perturbation sequences generated from each ImageNet validation image. Each sequence contains more than 30 frames, so to counteract an increase in dataset size and evaluation time only 10 common perturbations are used.

This dataset is used for anonymous data sharing for our NeurIPS 2025 submission.

The dataset used in the paper is a mixture of Gaussians, CIFAR-10, STL-10, CelebA, and ImageNet.

In addition to maintaining low carbon footprints, Ours method performs similarly to the KD [26] method.

Spiking Neural Networks (SNNs) have recently emerged as an alternative to deep learning owing to sparse, asynchronous and binary event (or spike) driven processing, that can yield huge energy efficiency benefits on neuromorphic hardware. However, SNNs convey temporally-varying spike activation through time that is likely to induce a large variation of forward activation and backward gradients, resulting in unstable training. To address this training issue in SNNs, we revisit Batch Normalization (BN) and propose a temporal Batch Normalization Through Time (BNTT) technique. Different from previous BN techniques with SNNs, we find that varying the BN parameters at every time-step allows the model to learn the time-varying input distribution better. Specifically, our proposed BNTT decouples the parameters in a BNTT layer along the time axis to capture the temporal dynamics of spikes. We demonstrate BNTT on CIFAR-10, CIFAR-100, Tiny-ImageNet, event-driven DVS-CIFAR10 datasets, and Sequential MNIST and show near state-of-the-art performance. We conduct comprehensive analysis on the temporal characteristic of BNTT and showcase interesting benefits toward robustness against random and adversarial noise. Further, by monitoring the learnt parameters of BNTT, we find that we can do temporal early exit. That is, we can reduce the inference latency by ~5 − 20 time-steps from the original training latency. The code has been released at https://github.com/Intelligent-Computing-Lab-Yale/BNTT-Batch-Normalization-Through-Time.

The PatchCamelyon benchmark is a new and challenging image classification dataset. It consists of 327.680 color images (96 x 96px) extracted from histopathologic scans of lymph node sections. Each image is annoted with a binary label indicating presence of metastatic tissue. PCam provides a new benchmark for machine learning models: bigger than CIFAR10, smaller than imagenet, trainable on a single GPU. ## Why PCam Fundamental machine learning advancements are predominantly evaluated on straight-forward natural-image classification datasets. Think MNIST, CIFAR, SVHN. Medical imaging is becoming one of the major applications of ML and we believe it deserves a spot on the list of go-to ML datasets. Both to challenge future work, and to steer developments into directions that are beneficial for this domain. We think PCam can play a role in this. It packs the clinically-relevant task of metastasis detection into a straight-forward binary image classification task, akin to CIFAR-10 and MNIST

The dataset used in the paper is not explicitly described, but it is mentioned that the authors used the CLIP model and the CIFAR-10 and ImageNet datasets.

The dataset used in the paper is not explicitly described, but it is mentioned that the authors used the CIFAR-10 and ImageNet-2012 datasets.

DenseNet-161

Densely Connected Convolutional Networks

Recent work has shown that convolutional networks can be substantially deeper, more accurate, and efficient to train if they contain shorter connections between layers close to the input and those close to the output. In this paper, we embrace this observation and introduce the Dense Convolutional Network (DenseNet), which connects each layer to every other layer in a feed-forward fashion. Whereas traditional convolutional networks with L layers have L connections - one between each layer and its subsequent layer - our network has L(L+1)/2 direct connections. For each layer, the feature-maps of all preceding layers are used as inputs, and its own feature-maps are used as inputs into all subsequent layers. DenseNets have several compelling advantages: they alleviate the vanishing-gradient problem, strengthen feature propagation, encourage feature reuse, and substantially reduce the number of parameters. We evaluate our proposed architecture on four highly competitive object recognition benchmark tasks (CIFAR-10, CIFAR-100, SVHN, and ImageNet). DenseNets obtain significant improvements over the state-of-the-art on most of them, whilst requiring less memory and computation to achieve high performance. Code and models are available at this https URL.

Authors: Gao Huang, Zhuang Liu, Kilian Q. Weinberger, Laurens van der Maaten
https://arxiv.org/abs/1608.06993

https://imgur.com/wWHWbQt.jpg" alt="DenseNet">

DenseNet Architectures

https://imgur.com/oiTdqJL.jpg" alt="DenseNet Architectures">

What is a Pre-trained Model?

A pre-trained model has been previously trained on a dataset and contains the weights and biases that represent the features of whichever dataset it was trained on. Learned features are often transferable to different data. For example, a model trained on a large dataset of bird images will contain learned features like edges or horizontal lines that you would be transferable your dataset.

Why use a Pre-trained Model?

Pre-trained models are beneficial to us for many reasons. By using a pre-trained model you are saving time. Someone else has already spent the time and compute resources to learn a lot of features and your model will likely benefit from it.

Adversarial training has become a primary method for enhancing the robustness of deep learning models. In recent years, fast adversarial training methods have gained widespread attention due to their lower computational cost. However, since fast adversarial training uses single-step adversarial attacks instead of multi-step attacks, the generated adversarial examples lack diversity, making models prone to catastrophic overfitting and loss of robustness. Existing methods to prevent catastrophic overfitting have certain shortcomings, such as poor robustness due to insufficient strength of generated adversarial examples, and low accuracy caused by excessive total perturbation. To address these issues, this paper proposes a fast adversarial training method—fast adversarial training with adaptive similarity step size (ATSS). In this method, random noise is first added to the input clean samples, and the model then calculates the gradient for each input sample. The perturbation step size for each sample is determined based on the similarity between the input noise and the gradient direction. Finally, adversarial examples are generated based on the step size and gradient for adversarial training. We conduct various adversarial attack tests on ResNet18 and VGG19 models using the CIFAR-10, CIFAR-100 and Tiny ImageNet datasets. The experimental results demonstrate that our method effectively avoids catastrophic overfitting. And compared to other fast adversarial training methods, ATSS achieves higher robustness accuracy and clean accuracy, with almost no additional training cost.

The dataset used in the paper is a de-noising diffusion probabilistic model (DDPM) trained on CIFAR-10, CIFAR-100, STL-10, and Tiny-ImageNet.

Results with * are rerun using our training strategy, while those without * are the results reported in their original papers [17, 18, 20]. β represents the gradient shrinking factor used in FDG [21] and γ denotes the learning rate shrinking factors in DTRP.