Ensuring the overall end-user experience is a challenging task in arbitrary style transfer (AST) due to the subjective nature of style transfer quality. A good practice is to provide users many instead of one AST result. However, existing approaches require to run multiple AST models or inference a diversified AST (DAST) solution multiple times, and thus they are either slow in speed or limited in diversity. In this paper, we propose a novel solution ensuring both efficiency and diversity for generating multiple user-controllable AST results by systematically modulating AST behavior at run-time. We begin with reformulating three prominent AST methods into a unified assign-and-mix problem and discover that the entropies of their assignment matrices exhibit a large variance. We then solve the unified problem in an optimal transport framework using the Sinkhorn-Knopp algorithm with a user input ε to control the said entropy and thus modulate stylization. Empirical results demonstrate the superiority of the proposed solution, with speed and stylization quality comparable to or better than existing AST and significantly more diverse than previous DAST works. Code is available at https://github.com/cplusx/eps-Assign-and-Mix.

Multi-turn textual feedback-based fashion image retrieval focuses on a real-world setting, where users can iteratively provide information to refine retrieval results until they find an item that fits all their requirements. In this work, we present a novel memory-based method, called FashionNTM, for such a multi-turn system. Our framework incorporates a new Cascaded Memory Neural Turing Machine (CM-NTM) approach for implicit state management, thereby learning to integrate information across all past turns to retrieve new images, for a given turn. Unlike vanilla Neural Turing Machine (NTM), our CM-NTM operates on multiple inputs, which interact with their respective memories via individual read and write heads, to learn complex relationships. Extensive evaluation results show that our proposed method outperforms the previous state-of-the-art algorithm by 50.5%, on Multi-turn FashionIQ [60] – the only existing multi-turn fashion dataset currently, in addition to having a relative improvement of 12.6% on Multi-turn Shoes – an extension of the singleturn Shoes dataset [5] that we created in this work. Further analysis of the model in a real-world interactive setting demonstrates two important capabilities of our model – memory retention across turns, and agnosticity to turn order for non-contradictory feedback. Finally, user study results show that images retrieved by FashionNTM were favored by 83.1% over other multi-turn models.

Fashion image retrieval based on a query pair of reference image and natural language feedback is a challenging task that requires models to assess fashion related information from visual and textual modalities simultaneously. We propose a new vision-language transformer based model, FashionVLP, that brings the prior knowledge contained in large image-text corpora to the domain of fashion image retrieval, and combines visual information from multiple levels of context to effectively capture fashion-related information. While queries are encoded through the transformer layers, our asymmetric design adopts a novel attention-based approach for fusing target image features without involving text or transformer layers in the process. Extensive results show that FashionVLP achieves the state-of-the-art performance on benchmark datasets, with a large 23% relative improvement on the challenging FashionIQ dataset, which contains complex natural language feedback.

Neural Style Transfer (NST) has quickly evolved from single-style to infinite-style models, also known as Arbitrary Style Transfer (AST). Although appealing results have been widely reported in literature, our empirical studies on four well-known AST approaches (GoogleMagenta [14], AdaIN [19], LinearTransfer [29], and SANet [37]) show that more than 50% of the time, AST stylized images are not acceptable to human users, typically due to under- or over-stylization. We systematically study the cause of this imbalanced style transferability (IST ) and propose a simple yet effective solution to mitigate this issue. Our studies show that the IST issue is related to the conventional AST style loss, and reveal that the root cause is the equal weightage of training samples irrespective of the properties of their corresponding style images, which biases the model towards certain styles. Through investigation of the theoretical bounds of the AST style loss, we propose a new loss that largely overcomes IST . Theoretical analysis and experimental results validate the effectiveness of our loss, with over 80% relative improvement in style deception rate and 98% relatively higher preference in human evaluation.

We present a privacy preserving machine learning method for images that separates task-relevant information from task-irrelevant information. Our primary hypothesis is that by revealing the minimal number of pixels required for a task we can provide the most privacy preserving guarantees. Specifically, we propose an adversarial method that masks out task-irrelevant information from an image for preserving privacy. The proposed method only uses task-specific label information and no privacy annotations such as identity of the subject, gender, race, etc., are required. We validate the proposed method on face attribute prediction on the CelebA dataset and emotion recognition on the FER+ dataset, showing that we can preserve visual privacy with little degradation in the task performance.

Object detection models perform well at localizing and classifying objects that they are shown during training. However, due to the difficulty and cost associated with creating and annotating detection datasets, trained models detect a limited number of object types with unknown objects treated as background content. This hinders the adoption of conventional detectors in real-world applications like large-scale object matching, visual grounding, visual relation prediction, obstacle detection (where it is more important to determine the presence and location of objects than to find specific types), etc. We propose class-agnostic object detection as a new problem that focuses on detecting objects irrespective of their object-classes. Specifically, the goal is to predict bounding boxes for all objects in an image but not their object-classes. The predicted boxes can then be consumed by another system to perform application-specific classification, retrieval, etc. We propose training and eval uation protocols for benchmarking class-agnostic detectors to advance future research in this domain. Finally, we propose (1) baseline methods and (2) a new adversarial learning framework for class-agnostic detection that forces the model to exclude class-specific information from features used for predictions. Experimental results show that adversarial learning improves class-agnostic detection efficacy.

We propose a novel approach to achieving invariance for deep neural networks in the form of inducing amnesia to unwanted factors of data through a new adversarial forgetting mechanism. We show that the forgetting mechanism serves as an information-bottleneck, which is manipulated by the adversarial training to learn invariance to unwanted factors. Empirical results show that the proposed framework achieves state-of-the-art performance at learning invariance in both nuisance and bias settings on a diverse collection of datasets and tasks.

Fake news often involves semantic manipulations across modalities such as image, text, location etc and requires the development of multimodal semantic forensics for its detection. Recent research has centered the problem around images, calling it image repurposing - where a digitally unmanipulated image is semantically misrepresented by means of its accompanying multimodal metadata such as captions, location, etc. The image and metadata together comprise a multimedia package. The problem setup requires algorithms to perform multimodal semantic forensics to authenticate a query multimedia package using a reference dataset of potentially related packages as evidences. Existing methods are limited to using a single evidence (retrieved package), which ignores potential performance improvement from the use of multiple evidences. In this work, we introduce a novel graph neural network based model for multimodal semantic forensics, which effectively utilizes multiple retrieved packages as evidences and is scalable with the number of evidences. We compare the scalability and performance of our model against existing methods. Experimental results show that the proposed model outperforms existing state-of-the-art algorithms with an error reduction of up to 25 %.

Image repurposing is a commonly used method for spreading misinformation on social media and online forums, which involves publishing untampered images with modified metadata to create rumors and further propaganda. While manual verification is possible, given vast amounts of verified knowledge available on the internet, the increasing prevalence and ease of this form of semantic manipulation call for the development of robust automatic ways of assessing the semantic integrity of multimedia data. In this paper, we present a novel method for image repurposing detection that is based on the real-world adversarial interplay between a bad actor who repurposes images with counterfeit metadata and a watchdog who verifies the semantic consistency between images and their accompanying metadata, where both players have access to a reference dataset of verified content, which they can use to achieve their goals. The proposed method exhibits state-of-the-art performance on location-identity, subject-identity and painting-artist verification, showing its efficacy across a diverse set of scenarios.

Deep neural network models for speech recognition have achieved great success recently, but they can learn incorrect associations between the target and nuisance factors of speech (e.g., speaker identities, background noise, etc.), which can lead to overfitting. While several methods have been proposed to tackle this problem, existing methods incorporate additional information about nuisance factors during training to develop invariant models. However, enumeration of all possible nuisance factors in speech data and the collection of their annotations is difficult and expensive. We present a robust training scheme for end-to-end speech recognition that adopts an unsupervised adversarial invariance induction framework to separate out essential factors for speech-recognition from nuisances without using any supplementary labels besides the transcriptions. Experiments show that the speech recognition model trained with the proposed training scheme achieves relative improvements of 5.48% on WSJ0, 6.16% on CHiME3, and 6.61% on TIMIT dataset over the base model. Additionally, the proposed method achieves a relative improvement of 14.44% on the combined WSJ0+CHiME3 dataset.

Data representations that contain all the information about target variables but are invariant to nuisance factors benefit supervised learning algorithms by preventing them from learning associations between these factors and the targets, thus reducing overfitting. We present a novel unsupervised invariance induction framework for neural networks that learns a split representation of data through competitive training between the prediction task and a reconstruction task coupled with disentanglement, without needing any labeled information about nuisance factors or domain knowledge. We describe an adversarial instantiation of this framework and provide analysis of its working. Our unsupervised model outperforms state-of-the-art methods, which are supervised, at inducing invariance to inherent nuisance factors, effectively using synthetic data augmentation to learn invariance, and domain adaptation. Our method can be applied to any prediction task, eg., binary/multi-class classification or regression, without loss of generality.

Real-world multimedia data is often composed of multiple modalities such as an image or a video with associated text (e.g., captions, user comments, etc.) and metadata. Such multimodal data packages are prone to manipulations, where a subset of these modalities can be altered to misrepresent or repurpose data packages, with possible malicious intent. It is therefore important to develop methods to assess or verify the integrity of these multimedia packages. Using computer vision and natural language processing methods to directly compare the image (or video) and the associated caption to verify the integrity of a media package is only possible for a limited set of objects and scenes. In this paper we present a novel deep-learning-based approach that uses a reference set of multimedia packages to assess the semantic integrity of multimedia packages containing images and captions. We construct a joint embedding of images and captions with deep multimodal representation learning on the reference dataset in a framework that also provides image-caption consistency scores (ICCSs). The integrity of query media packages is assessed as the inlierness of the query ICCSs with respect to the reference dataset. We present the MultimodAl Information Manipulation dataset (MAIM), a new dataset of media packages from Flickr, which we are making available to the research community. We use both the newly created dataset as well as Flickr30K and MS COCO datasets to quantitatively evaluate our proposed approach. The reference dataset does not contain unmanipulated versions of tampered query packages. Our method is able to achieve F-1 scores of 0.75, 0.89 and 0.94 on MAIM, Flickr30K and MS COCO, respectively, for detecting semantically incoherent media packages.