Personalized outfit compatibility learning is an emerging yet challenging task. Most of the existing methods focus on general outfit compatibility learning. Although a few works have been proposed ...for personalized fashion compatibility, they either considered user preference on fashion items with specific patterns or design elements or recommended outfits based on the overall visual similarity according to the users’ preferred collections. This paper adopts physical and fashion attributes for effective personalized fashion compatibility evaluation and recommendation. The physical attributes are concluded into seven aspects: body shape, skin color, hairstyle, hair color, height, breast size (breasts), and color contrast. The personalized outfit compatibility problem in this paper is a multi-label classification problem and formulated as an optimization function with outfit images, fashion attributes, and physical attributes as input. It is the first attempt to solve the problem by discovering the correlation between visual image features, fashion attributes, and physical attributes. Specifically, the correlation is learned with two transformer encoders by updating attention weights of different embedding pairs during the training process. The model can not only predict the fashion attributes of the outfit’s top, bottom, shoes, and bag items, but also predict the incompatible physical attributes of an individual towards the given outfit. It can be used to recommend outfits that best fit an individual and the predicted fashion attributes can be used for result explanation. The O4U dataset, which contains rich annotations of fashion item attributes and human physical attributes of the outfits, is used to evaluate the performance of the proposed method. The quantitative and qualitative results show that the proposed method outperforms state-of-the-art methods for personalized outfit compatibility evaluation.
•We first consider individual physical attributes for personalized recommendation.•Correlation between visual features and physical attributes is explored.•Experimental results demonstrate the superiority of the proposed model.
Body shape is a crucial factor in outfit recommendation. Previous studies that directly used body measurement data to investigate the relationship between body shape and outfit have achieved limited ...performance due to oversimplified body shape representations. This paper proposes a Visual Body-shape-Aware Network (ViBA-Net) to improve the fashion compatibility model's awareness of human body shape through visual-level information. Specifically, ViBA-Net consists of three modules: a body-shape embedding module, which extracts visual and anthropometric features of body shape from a newly introduced large-scale body shape dataset; an outfit embedding module, which learns the outfit representation based on visual features extracted from a try-on image and textual features extracted from fashion attributes; and a joint embedding module, which jointly models the relationship between the representations of body shape and outfit. ViBA-Net is designed to generate attribute-level explanations for the evaluation results based on the computed attention weights. The effectiveness of ViBA-Net is evaluated on two mainstream datasets through qualitative and quantitative analysis. Data and code are released 1 .
The goal of this paper is to model the fashion compatibility of an outfit and provide the explanations. We first extract features of all attributes of all items via convolutional neural networks, and ...then train the bidirectional Long Short-term Memory (Bi-LSTM) model to learn the compatibility of an outfit by treating these attribute features as a sequence. Gradient penalty regularization is exploited for training inter-factor compatibility net which is used to compute the loss for judgment and provide its explanation which is generated from the recognized reasons related to the judgment. To train and evaluate the proposed approach, we expanded the EVALUATION3 dataset in terms of the number of items and attributes. Experiment results show that our approach can successfully evaluate compatibility with reason.
How Good Is Aesthetic Ability of a Fashion Model? Zou, Xingxing; Pang, Kaicheng; Zhang, Wen ...
2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR),
2022-June
Conference Proceeding
We introduce A100 (Aesthetic 100) to assess the aesthetic ability of the fashion compatibility models. To date, it is the first work to address the AI model's aesthetic ability with detailed ...characterization based on the professional fashion domain knowledge. A100 has several desirable characteristics: 1. Completeness. It covers all types of standards in the fashion aesthetic system through two tests, namely LAT (Liberalism Aesthetic Test) and AAT (Academicism Aesthetic Test); 2. Reliability. It is training data agnostic and consistent with major indicators. It provides a fair and objective judgment for model comparison. 3. Explainability. Better than all previous indicators, the A100 further identifies essential characteristics of fashion aesthetics, thus showing the model's performance on more fine-grained dimensions, such as Color, Balance, Material, etc. Experimental results prove the advance of the A100 in the aforementioned aspects. All data can be found at https://github.com/AemikaChow/AiDLab-fAshIon-Data.
Fashion compatibility models enable online retailers to easily obtain a large number of outfit compositions with good quality. However, effective fashion recommendation demands precise service for ...each customer with a deeper cognition of fashion. In this paper, we conduct the first study on fashion cognitive learning, which is fashion recommendations conditioned on personal physical information. To this end, we propose a Fashion Cognitive Network (FCN) to learn the relationships among visual-semantic embedding of outfit composition and appearance features of individuals. FCN contains two submodules, namely outfit encoder and Multi-label Graph Neural Network (ML-GCN). The outfit encoder uses a convolutional layer to encode an outfit into an outfit embedding. The latter module learns label classifiers via stacked GCN. We conducted extensive experiments on the newly collected O4U dataset, and the results provide strong qualitative and quantitative evidence that our framework outperforms alternative methods.
Triple-negative breast cancer (TNBC) causes great suffering to patients because of its heterogeneity, poor prognosis, and chemotherapy resistance. Ferroptosis is characterized by iron-dependent ...oxidative damage by accumulating intracellular lipid peroxides to lethal levels, and plays a vital role in the treatment of TNBC based on its intrinsic characteristics. To identify the relationship between chemotherapy resistance and ferroptosis in TNBC, we analyzed the single cell RNA-sequencing public dataset of GSE205551. It was found that the expression of Gpx4 in DOX-resistant TNBC cells was significantly higher than that in DOX-sensitive TNBC cells. Based on this finding, we hypothesize that inducing ferroptosis by inhibiting the expression of Gpx4 can reduce the resistance of TNBC to DOX and enhance the therapeutic effect of chemotherapy on TNBC. Herein, dihydroartemisinin (DHA)-loaded polyglutamic acid-stabilized Fe
O
magnetic nanoparticles (Fe
O
-PGA-DHA) was combined with DOX-loaded polyaspartic acid-stabilized Fe
O
magnetic nanoparticles (Fe
O
-PASP-DOX) for ferroptosis-enhanced chemotherapy of TNBC. Compared with Fe
O
-PASP-DOX, Fe
O
-PGA-DHA + Fe
O
-PASP-DOX demonstrated significantly stronger cytotoxicity against different TNBC cell lines and achieved significantly more intracellular accumulation of reactive oxygen species and lipid peroxides. Furthermore, transcriptomic analyses demonstrated that Fe
O
-PASP-DOX-induced apoptosis could be enhanced by Fe
O
-PGA-DHA-induced ferroptosis and Fe
O
-PGA-DHA + Fe
O
-PASP-DOX might trigger ferroptosis in MDA-MB-231 cells by inhibiting the PI3K/AKT/mTOR/GPX4 pathway. Fe
O
-PGA-DHA + Fe
O
-PASP-DOX showed superior anti-tumor efficacy on MDA-MB-231 tumor-bearing mice, providing great potential for improving the therapeutic effect of TNBC.
Under the dual carbon strategy, with the frequent occurrence of extreme weather and the further increase in uncertainty of multi-user behavior, it is urgent to improve the stability of the heating ...systems and reduce heating energy consumption. Aiming at the problem of fault-disturbance control of the multi-user heating network in an integrated energy system, this paper proposes a novel analysis method of resistance–capacitance reactance based on the circuit principle to construct a dynamic thermal-power-flow model of the whole link of the multi-user heating network and analyze the fault-disturbance propagation characteristics of the heating network by this model. It shows that the difference in disturbance characteristics of different users in a multi-user heating network mainly depends on the characteristics of the heating pipeline between the heat user and the heat source, which provides a necessary basis for formulating intelligent control strategies against fault disturbance. Finally, taking a typical daily outdoor temperature in Beijing in winter as an example, this paper compares two different heating strategies and the blocker installation methods of the multi-user heating network to obtain a better heating strategy under actual conditions. Considering the heating fault disturbance, this paper proposes a novel intelligent heating strategy whose heating temperature will decrease during the fault-disturbance time, with an energy saving of about 16.5% compared with the heating strategy under actual conditions during the same period.