Olfaction -- how molecules are perceived as odors to humans -- remains poorly understood. Recently, the principal odor map (POM) was introduced to digitize the olfactory properties of single compounds. However, smells in real life are not pure single molecules, but complex mixtures of molecules, whose representations remain relatively under-explored. In this work, we introduce POMMix, an extension of the POM to represent mixtures. Our representation builds upon the symmetries of the problem space in a hierarchical manner: (1) graph neural networks for building molecular embeddings, (2) attention mechanisms for aggregating molecular representations into mixture representations, and (3) cosine prediction heads to encode olfactory perceptual distance in the mixture embedding space. POMMix achieves state-of-the-art predictive performance across multiple datasets. We also evaluate the generalizability of the representation on multiple splits when applied to unseen molecules and mixture sizes. Our work advances the effort to digitize olfaction, and highlights the synergy of domain expertise and deep learning in crafting expressive representations in low-data regimes.
We consider the problem of joint source-channel coding for semantic communication from a rateless perspective, the purpose of which is to settle the balance between reliability (distortion/perception) and effectiveness (rate) of transmission over uncertain channels. In particular, we propose a more general communication objective that minimizes the perceptual distance by incorporating a semantic-level reconstruction objective in addition to the conventional pixel-level reconstruction objective. Based on the proposed objective, we then propose a novel JSCC coding scheme called rateless stochastic coding (RSC) by introducing a generative decoder and dithered quantization. The coding scheme enables reconstruction based on both distortion and perception metrics through rateless transmission. Extensive experiments demonstrate that the proposed RSC can achieve variable rates of transmission maintaining an excellent trade-off between distortion and perception.
Diffusion models are the state of the art in text-to-image generation, but their perceptual variability remains understudied. In this paper, we examine how prompts affect image variability in black-box diffusion-based models. We propose W1KP, a human-calibrated measure of variability in a set of images, bootstrapped from existing image-pair perceptual distances. Current datasets do not cover recent diffusion models, thus we curate three test sets for evaluation. Our best perceptual distance outperforms nine baselines by up to 18 points in accuracy, and our calibration matches graded human judgements 78% of the time. Using W1KP, we study prompt reusability and show that Imagen prompts can be reused for 10-50 random seeds before new images become too similar to already generated images, while Stable Diffusion XL and DALL-E 3 can be reused 50-200 times. Lastly, we analyze 56 linguistic features of real prompts, finding that the prompt's length, CLIP embedding norm, concreteness, and word senses influence variability most. As far as we are aware, we are the first to analyze diffusion variability from a visuolinguistic perspective. Our project page is at http://w1kp.com.
The two-alternative forced choice (2AFC) experimental method is popular in the visual perception literature, where practitioners aim to understand how human observers perceive distances within triplets made of a reference image and two distorted versions. In the past, this had been conducted in controlled environments, with triplets sharing images, so it was possible to rank the perceived quality. This ranking would then be used to evaluate perceptual distance models against the experimental data. Recently, crowd-sourced perceptual datasets have emerged, with no images shared between triplets, making ranking infeasible. Evaluating perceptual distance models using this data reduces the judgements on a triplet to a binary decision, namely, whether the distance model agrees with the human decision - which is suboptimal and prone to misleading conclusions. Instead, we statistically model the underlying decision-making process during 2AFC experiments using a binomial distribution. Having enough empirical data, we estimate a smooth and consistent distribution of the judgements on the reference-distorted distance plane, according to each distance model. By applying maximum likelihood, we estimate the parameter of the local binomial distribution, and a global measurement of the expected log-likelihood of the measured responses. We calculate meaningful and well-founded metrics for the distance model, beyond the mere prediction accuracy as percentage agreement, even with variable numbers of judgements per triplet -- key advantages over both classical and neural network methods.
The volume of image repositories continues to grow. Despite the availability of content-based addressing, we still lack a lightweight tool that allows us to discover images of distinct characteristics from a large collection. In this paper, we propose a fast and training-free algorithm for novel image discovery. The key of our algorithm is formulating a collection of images as a perceptual distance-weighted graph, within which our task is to locate the K-densest subgraph that corresponds to a subset of the most unique images. While solving this problem is not just NP-hard but also requires a full computation of the potentially huge distance matrix, we propose to relax it into a K-sparse eigenvector problem that we can efficiently solve using stochastic gradient descent (SGD) without explicitly computing the distance matrix. We compare our algorithm against state-of-the-arts on both synthetic and real datasets, showing that it is considerably faster to run with a smaller memory footprint while able to mine novel images more accurately.
We propose an end-to-end learned image compression codec wherein the analysis transform is jointly trained with an object classification task. This study affirms that the compressed latent representation can predict human perceptual distance judgments with an accuracy comparable to a custom-tailored DNN-based quality metric. We further investigate various neural encoders and demonstrate the effectiveness of employing the analysis transform as a perceptual loss network for image tasks beyond quality judgments. Our experiments show that the off-the-shelf neural encoder proves proficient in perceptual modeling without needing an additional VGG network. We expect this research to serve as a valuable reference developing of a semantic-aware and coding-efficient neural encoder.
Novelty detection is a fundamental task of machine learning which aims to detect abnormal ($\textit{i.e.}$ out-of-distribution (OOD)) samples. Since diffusion models have recently emerged as the de facto standard generative framework with surprising generation results, novelty detection via diffusion models has also gained much attention. Recent methods have mainly utilized the reconstruction property of in-distribution samples. However, they often suffer from detecting OOD samples that share similar background information to the in-distribution data. Based on our observation that diffusion models can \emph{project} any sample to an in-distribution sample with similar background information, we propose \emph{Projection Regret (PR)}, an efficient novelty detection method that mitigates the bias of non-semantic information. To be specific, PR computes the perceptual distance between the test image and its diffusion-based projection to detect abnormality. Since the perceptual distance often fails to capture semantic changes when the background information is dominant, we cancel out the background bias by comparing it against recursive projections. Extensive experiments demonstrate that PR outperforms the prior art of generative-model-based novelty detection methods by a significant margin.
Styled Handwritten Text Generation (Styled HTG) is an important task in document analysis, aiming to generate text images with the handwriting of given reference images. In recent years, there has been significant progress in the development of deep learning models for tackling this task. Being able to measure the performance of HTG models via a meaningful and representative criterion is key for fostering the development of this research topic. However, despite the current adoption of scores for natural image generation evaluation, assessing the quality of generated handwriting remains challenging. In light of this, we devise the Handwriting Distance (HWD), tailored for HTG evaluation. In particular, it works in the feature space of a network specifically trained to extract handwriting style features from the variable-lenght input images and exploits a perceptual distance to compare the subtle geometric features of handwriting. Through extensive experimental evaluation on different word-level and line-level datasets of handwritten text images, we demonstrate the suitability of the proposed HWD as a score for Styled HTG. The pretrained model used as backbone will be released to ease the adoption of the score, aiming to provide a valuable tool for evaluating HTG models and thus contributing to advancing this important research area.
Deep neural networks are known to be vulnerable to adversarial perturbations. The amount of these perturbations are generally quantified using $L_p$ metrics, such as $L_0$, $L_2$ and $L_\infty$. However, even when the measured perturbations are small, they tend to be noticeable by human observers since $L_p$ distance metrics are not representative of human perception. On the other hand, humans are less sensitive to changes in colorspace. In addition, pixel shifts in a constrained neighborhood are hard to notice. Motivated by these observations, we propose a method that creates adversarial examples by applying spatial transformations, which creates adversarial examples by changing the pixel locations independently to chrominance channels of perceptual colorspaces such as $YC_{b}C_{r}$ and $CIELAB$, instead of making an additive perturbation or manipulating pixel values directly. In a targeted white-box attack setting, the proposed method is able to obtain competitive fooling rates with very high confidence. The experimental evaluations show that the proposed method has favorable results in terms of approximate perceptual distance between benign and adversarially generated images. The source code is publicly available at https://github.com/ayberkydn/stadv-torch
Lossy image coding standards such as JPEG and MPEG have successfully achieved high compression rates for human consumption of multimedia data. However, with the increasing prevalence of IoT devices, drones, and self-driving cars, machines rather than humans are processing a greater portion of captured visual content. Consequently, it is crucial to pursue an efficient compressed representation that caters not only to human vision but also to image processing and machine vision tasks. Drawing inspiration from the efficient coding hypothesis in biological systems and the modeling of the sensory cortex in neural science, we repurpose the compressed latent representation to prioritize semantic relevance while preserving perceptual distance. Our proposed method, Compressed Perceptual Image Patch Similarity (CPIPS), can be derived at a minimal cost from a learned neural codec and computed significantly faster than DNN-based perceptual metrics such as LPIPS and DISTS.
We propose a method to estimate the mechanical parameters of fabrics using a casual capture setup with a depth camera. Our approach enables to create mechanically-correct digital representations of real-world textile materials, which is a fundamental step for many interactive design and engineering applications. As opposed to existing capture methods, which typically require expensive setups, video sequences, or manual intervention, our solution can capture at scale, is agnostic to the optical appearance of the textile, and facilitates fabric arrangement by non-expert operators. To this end, we propose a sim-to-real strategy to train a learning-based framework that can take as input one or multiple images and outputs a full set of mechanical parameters. Thanks to carefully designed data augmentation and transfer learning protocols, our solution generalizes to real images despite being trained only on synthetic data, hence successfully closing the sim-to-real loop.Key in our work is to demonstrate that evaluating the regression accuracy based on the similarity at parameter space leads to an inaccurate distances that do not match the human perception. To overcome this, we propose a novel metric for fabric drape similarity that operates on the image domain instead on the parameter space, allowing us to evaluate our estimation within the context of a similarity rank. We show that out metric correlates with human judgments about the perception of drape similarity, and that our model predictions produce perceptually accurate results compared to the ground truth parameters.
This paper develops a unified framework for image-to-image translation based on conditional diffusion models and evaluates this framework on four challenging image-to-image translation tasks, namely colorization, inpainting, uncropping, and JPEG restoration. Our simple implementation of image-to-image diffusion models outperforms strong GAN and regression baselines on all tasks, without task-specific hyper-parameter tuning, architecture customization, or any auxiliary loss or sophisticated new techniques needed. We uncover the impact of an L2 vs. L1 loss in the denoising diffusion objective on sample diversity, and demonstrate the importance of self-attention in the neural architecture through empirical studies. Importantly, we advocate a unified evaluation protocol based on ImageNet, with human evaluation and sample quality scores (FID, Inception Score, Classification Accuracy of a pre-trained ResNet-50, and Perceptual Distance against original images). We expect this standardized evaluation protocol to play a role in advancing image-to-image translation research. Finally, we show that a generalist, multi-task diffusion model performs as well or better than task-specific specialist counterparts. Check out https://diffusion-palette.github.io for an overview of the results.
Generative adversarial networks (GANs) have proven to be surprisingly efficient for image editing by inverting and manipulating the latent code corresponding to a natural image. This property emerges from the disentangled nature of the latent space. In this paper, we identify two geometric limitations of such latent space: (a) euclidean distances differ from image perceptual distance, and (b) disentanglement is not optimal and facial attribute separation using linear model is a limiting hypothesis. We thus propose a new method to learn a proxy latent representation using normalizing flows to remedy these limitations, and show that this leads to a more efficient space for face image editing.
It has been demonstrated many times that the behavior of the human visual system is connected to the statistics of natural images. Since machine learning relies on the statistics of training data as well, the above connection has interesting implications when using perceptual distances (which mimic the behavior of the human visual system) as a loss function. In this paper, we aim to unravel the non-trivial relationships between the probability distribution of the data, perceptual distances, and unsupervised machine learning. To this end, we show that perceptual sensitivity is correlated with the probability of an image in its close neighborhood. We also explore the relation between distances induced by autoencoders and the probability distribution of the training data, as well as how these induced distances are correlated with human perception. Finally, we find perceptual distances do not always lead to noticeable gains in performance over Euclidean distance in common image processing tasks, except when data is scarce and the perceptual distance provides regularization. We propose this may be due to a \emph{double-counting} effect of the image statistics, once in the perceptual distance and once in the training procedure.
Capturing interpretable variations has long been one of the goals in disentanglement learning. However, unlike the independence assumption, interpretability has rarely been exploited to encourage disentanglement in the unsupervised setting. In this paper, we examine the interpretability of disentangled representations by investigating two questions: where to be interpreted and what to be interpreted? A latent code is easily to be interpreted if it would consistently impact a certain subarea of the resulting generated image. We thus propose to learn a spatial mask to localize the effect of each individual latent dimension. On the other hand, interpretability usually comes from latent dimensions that capture simple and basic variations in data. We thus impose a perturbation on a certain dimension of the latent code, and expect to identify the perturbation along this dimension from the generated images so that the encoding of simple variations can be enforced. Additionally, we develop an unsupervised model selection method, which accumulates perceptual distance scores along axes in the latent space. On various datasets, our models can learn high-quality disentangled representations without supervision, showing the proposed modeling of interpretability is an effective proxy for achieving unsupervised disentanglement.
Our sensory systems transform external signals into neural activity, thereby producing percepts. We are endowed with an intuitive notion of similarity between percepts, that need not reflect the proximity of the physical properties of the corresponding external stimuli. The quantitative characterization of the geometry of percepts is therefore an endeavour that must be accomplished behaviorally. Here we characterized the geometry of color space using discrimination and matching experiments. We proposed an individually tailored metric defined in terms of the minimal chromatic difference required for each observer to differentiate a stimulus from its surround. Next, we showed that this perceptual metric was particularly adequate to describe two additional experiments, since it revealed the natural symmetry of perceptual computations. In one of the experiments, observers were required to discriminate two stimuli surrounded by a chromaticity that differed from that of the tested stimuli. In the perceptual coordinates, the change in discrimination thresholds induced by the surround followed a simple law that only depended on the perceptual distance between the surround and each of the two compared stimuli. In the other experiment, subjects were asked to match the color of two stimuli surrounded by two different chromaticities. Again, in the perceptual coordinates the induction effect produced by surrounds followed a simple, symmetric law. We conclude that the individually-tailored notion of perceptual distance reveals the symmetry of the laws governing perceptual computations.
We present a neural optimization model trained with reinforcement learning to solve the coordinate ordering problem for sets of star glyphs. Given a set of star glyphs associated to multiple class labels, we propose to use shape context descriptors to measure the perceptual distance between pairs of glyphs, and use the derived silhouette coefficient to measure the perception of class separability within the entire set. To find the optimal coordinate order for the given set, we train a neural network using reinforcement learning to reward orderings with high silhouette coefficients. The network consists of an encoder and a decoder with an attention mechanism. The encoder employs a recurrent neural network (RNN) to encode input shape and class information, while the decoder together with the attention mechanism employs another RNN to output a sequence with the new coordinate order. In addition, we introduce a neural network to efficiently estimate the similarity between shape context descriptors, which allows to speed up the computation of silhouette coefficients and thus the training of the axis ordering network. Two user studies demonstrate that the orders provided by our method are preferred by users for perceiving class separation. We tested our model on different settings to show its robustness and generalization abilities and demonstrate that it allows to order input sets with unseen data size, data dimension, or number of classes. We also demonstrate that our model can be adapted to coordinate ordering of other types of plots such as RadViz by replacing the proposed shape-aware silhouette coefficient with the corresponding quality metric to guide network training.
In this paper, we propose a framework called TrustMAE to address the problem of product defect classification. Instead of relying on defective images that are difficult to collect and laborious to label, our framework can accept datasets with unlabeled images. Moreover, unlike most anomaly detection methods, our approach is robust against noises, or defective images, in the training dataset. Our framework uses a memory-augmented auto-encoder with a sparse memory addressing scheme to avoid over-generalizing the auto-encoder, and a novel trust-region memory updating scheme to keep the noises away from the memory slots. The result is a framework that can reconstruct defect-free images and identify the defective regions using a perceptual distance network. When compared against various state-of-the-art baselines, our approach performs competitively under noise-free MVTec datasets. More importantly, it remains effective at a noise level up to 40% while significantly outperforming other baselines.
This paper uses an axiomatic foundation to create a new measure for the cost of learning that allows for multiple perceptual distances in a single choice environment so that some events can be harder to differentiate between than others. The new measure maintains the tractability of Shannon's classic measure but produces richer choice predictions and identifies a new form of informational bias significant for welfare and counterfactual analysis.
Constructing adversarial examples in a black-box threat model injures the original images by introducing visual distortion. In this paper, we propose a novel black-box attack approach that can directly minimize the induced distortion by learning the noise distribution of the adversarial example, assuming only loss-oracle access to the black-box network. The quantified visual distortion, which measures the perceptual distance between the adversarial example and the original image, is introduced in our loss whilst the gradient of the corresponding non-differentiable loss function is approximated by sampling noise from the learned noise distribution. We validate the effectiveness of our attack on ImageNet. Our attack results in much lower distortion when compared to the state-of-the-art black-box attacks and achieves $100\%$ success rate on InceptionV3, ResNet50 and VGG16bn. The code is available at https://github.com/Alina-1997/visual-distortion-in-attack.