The rapid growth of the online fashion market brought demands for innovative fashion services and commerce platforms. With the recent success of deep learning, many applications employ AI technologies such as visual search and recommender systems to provide novel and beneficial services. In this paper, we describe applied technologies for AI-driven fashion social networking service that incorporate fashion e-commerce. In the application, people can share and browse their outfit-of-the-day (OOTD) photos, while AI analyzes them and suggests similar style OOTDs and related products. To this end, we trained deep learning based AI models for fashion and integrated them to build a fashion visual search system and a recommender system for OOTD. With aforementioned technologies, the AI-driven fashion SNS platform, iTOO, has been successfully launched.
Personalized outfit recommendation has recently been in the spotlight with the rapid growth of the online fashion industry. However, recommending outfits has two significant challenges that should be addressed. The first challenge is that outfit recommendation often requires a complex and large model that utilizes visual information, incurring huge memory and time costs. One natural way to mitigate this problem is to compress such a cumbersome model with knowledge distillation (KD) techniques that leverage knowledge from a pretrained teacher model. However, it is hard to apply existing KD approaches in recommender systems (RS) to the outfit recommendation because they require the ranking of all possible outfits while the number of outfits grows exponentially to the number of consisting clothing items. Therefore, we propose a new KD framework for outfit recommendation, called False Negative Distillation (FND), which exploits false-negative information from the teacher model while not requiring the ranking of all candidates. The second challenge is that the explosive number of outfit candidates amplifying the data sparsity problem, often leading to poor outfit representation. To tackle this issue, inspired by the recent success of contrastive learning (CL), we introduce a CL framework for outfit representation learning with two proposed data augmentation methods. Quantitative and qualitative experiments on outfit recommendation datasets demonstrate the effectiveness and soundness of our proposed methods.
Although BERT and its variants have reshaped the NLP landscape, it still remains unclear how best to derive sentence embeddings from such pre-trained Transformers. In this work, we propose a contrastive learning method that utilizes self-guidance for improving the quality of BERT sentence representations. Our method fine-tunes BERT in a self-supervised fashion, does not rely on data augmentation, and enables the usual [CLS] token embeddings to function as sentence vectors. Moreover, we redesign the contrastive learning objective (NT-Xent) and apply it to sentence representation learning. We demonstrate with extensive experiments that our approach is more effective than competitive baselines on diverse sentence-related tasks. We also show it is efficient at inference and robust to domain shifts.
Detecting anomalies is one fundamental aspect of a safety-critical software system, however, it remains a long-standing problem. Numerous branches of works have been proposed to alleviate the complication and have demonstrated their efficiencies. In particular, self-supervised learning based methods are spurring interest due to their capability of learning diverse representations without additional labels. Among self-supervised learning tactics, contrastive learning is one specific framework validating their superiority in various fields, including anomaly detection. However, the primary objective of contrastive learning is to learn task-agnostic features without any labels, which is not entirely suited to discern anomalies. In this paper, we propose a task-specific variant of contrastive learning named masked contrastive learning, which is more befitted for anomaly detection. Moreover, we propose a new inference method dubbed self-ensemble inference that further boosts performance by leveraging the ability learned through auxiliary self-supervision tasks. By combining our models, we can outperform previous state-of-the-art methods by a significant margin on various benchmark datasets.
Image-to-image translation aims to learn a mapping between different groups of visually distinguishable images. While recent methods have shown impressive ability to change even intricate appearance of images, they still rely on domain labels in training a model to distinguish between distinct visual features. Such dependency on labels often significantly limits the scope of applications since consistent and high-quality labels are expensive. Instead, we wish to capture visual features from images themselves and apply them to enable realistic translation without human-generated labels. To this end, we propose an unsupervised image-to-image translation method based on contrastive learning. The key idea is to learn a discriminator that differentiates between distinctive styles and let the discriminator supervise a generator to transfer those styles across images. During training, we randomly sample a pair of images and train the generator to change the appearance of one towards another while keeping the original structure. Experimental results show that our method outperforms the leading unsupervised baselines in terms of visual quality and translation accuracy.
Adversarial training is a defense technique that improves adversarial robustness of a deep neural network (DNN) by including adversarial examples in the training data. In this paper, we identify an overlooked problem of adversarial training in that these adversarial examples often have different semantics than the original data, introducing unintended biases into the model. We hypothesize that such non-semantics-preserving (and resultingly ambiguous) adversarial data harm the robustness of the target models. To mitigate such unintended semantic changes of adversarial examples, we propose semantics-preserving adversarial training (SPAT) which encourages perturbation on the pixels that are shared among all classes when generating adversarial examples in the training stage. Experiment results show that SPAT improves adversarial robustness and achieves state-of-the-art results in CIFAR-10 and CIFAR-100.
We propose a novel method that enables us to determine words that deserve to be emphasized from written text in visual media, relying only on the information from the self-attention distributions of pre-trained language models (PLMs). With extensive experiments and analyses, we show that 1) our zero-shot approach is superior to a reasonable baseline that adopts TF-IDF and that 2) there exist several attention heads in PLMs specialized for emphasis selection, confirming that PLMs are capable of recognizing important words in sentences.
Zero-shot constituency parsing aims to extract parse trees from neural models such as pre-trained language models (PLMs) without further training or the need to train an additional parser. This paper improves upon existing zero-shot parsing paradigms by introducing a novel chart-based parsing method, showing gains in zero-shot parsing performance. Furthermore, we attempt to broaden the range of zero-shot parsing applications by examining languages other than English and by utilizing multilingual models, demonstrating that it is feasible to generate parse tree-like structures for sentences in eight other languages using our method.
Recent works have shown that generative data augmentation, where synthetic samples generated from deep generative models are used to augment the training dataset, benefit certain NLP tasks. In this work, we extend this approach to the task of dialog state tracking for goal-oriented dialogs. Since, goal-oriented dialogs naturally exhibit a hierarchical structure over utterances and related annotations, deep generative data augmentation for the task requires the generative model to be aware of the hierarchical nature. We propose the Variational Hierarchical Dialog Autoencoder (VHDA) for modeling complete aspects of goal-oriented dialogs, including linguistic features and underlying structured annotations, namely dialog acts and goals. We also propose two training policies to mitigate issues that arise with training VAE-based models. Experiments show that our hierarchical model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialog state trackers, ultimately improving the dialog state tracking performances on various dialog domains. Surprisingly, the ability to jointly generate dialog features enables our model to outperform previous state-of-the-arts in related subtasks, such as language generation and user simulation.
With the recent success and popularity of pre-trained language models (LMs) in natural language processing, there has been a rise in efforts to understand their inner workings. In line with such interest, we propose a novel method that assists us in investigating the extent to which pre-trained LMs capture the syntactic notion of constituency. Our method provides an effective way of extracting constituency trees from the pre-trained LMs without training. In addition, we report intriguing findings in the induced trees, including the fact that pre-trained LMs outperform other approaches in correctly demarcating adverb phrases in sentences.