Reconstructing 3D assets from images, known as inverse rendering (IR), remains a challenging task due to its ill-posed nature. 3D Gaussian Splatting (3DGS) has demonstrated impressive capabilities for novel view synthesis (NVS) tasks. Methods apply it to relighting by separating radiance into BRDF parameters and lighting, yet produce inferior relighting quality with artifacts and unnatural indirect illumination due to the limited capability of each Gaussian, which has constant material parameters and normal, alongside the absence of physical constraints for indirect lighting. In this paper, we present a novel framework called Spatially-vayring Gaussian Inverse Rendering (SVG-IR), aimed at enhancing both NVS and relighting quality. To this end, we propose a new representation—Spatially-varying Gaussian (SVG)—that allows per-Gaussian spatially varying parameters. This enhanced representation is complemented by a SVG splatting scheme akin to vertex/fragment shading in traditional graphics pipelines. Furthermore, we integrate a physically-based indirect lighting model, enabling more realistic relighting. The proposed SVG-IR framework significantly improves rendering quality, outperforming state-of-the-art NeRF-based methods by 2.5 dB in peak signal-to-noise ratio (PSNR) and surpassing existing Gaussian-based techniques by 3.5 dB in relighting tasks, all while maintaining a real-time rendering speed. The source code is available at https://github.com/learner-shx/SVG-IR.

The limited spectrum resources have a particular impact on UAV-assisted semantic communication networks, which undoubtedly leads to poorer quality of service for users and inefficient communication. Therefore, a NOMA-based multi-UAVs assisted semantic cellular network framework is proposed in this paper, in which each UAV transmits semantic information to multiple users in the shared spectrum resource with different power using non-orthogonal multiple access transmission protocol, thereby achieving higher spectrum utilization. We optimize the quantity of semantic symbols, UAV trajectories, and power allocation concurrently to increase communication efficiency by maximizing the sum rate of semantic information transmission for all users. However, conventional convex optimization approaches have difficulty solving it due to the bi-directional mobility of UAVs and users. Therefore, an enhanced K-means algorithm is employed to create the relationship between UAVs and users periodically. Additionally, a deep reinforcement learning technique based on shared dueling double deep Q networks (SD3QN) is also presented to maximize the quantity of semantic symbols, 3D trajectories, and power allocation . Experimental results show that the proposed semantic cellular network achieves higher spectral efficiency . Meanwhile, the proposed algorithm can effectively reduce the training time and avoid the overestimation problem in Deep Q Networks (DQN). Furthermore, the suggested optimization strategy outperforms the benchmark schemes in terms of semantic sum rate.

Fig. 1.By modeling the correlation among input images with an encoder, together with an adapter-equipped decoder, our network achieves high-quality SVBRDF recovery on both isotropic and anisotropic (with roughness encoded in red and green channels) materials.Here we show re-rendered views for four materials under environment illumination.(Please use Adobe Acrobat and click the renderings to see the animation.)Capturing materials from the real world avoids laborious manual material authoring.However, recovering high-fidelity Spatially Varying Bidirectional Reflectance Distribution Function (SVBRDF) maps from a few captured images is challenging due to its ill-posed nature.Existing approaches have made extensive efforts to alleviate this ambiguity issue by leveraging generative models with latent space optimization or extracting features with variant encoder-decoders.Albeit the rendered images at input views can match input images, the problematic decomposition among maps leads to significant differences when rendered under novel views/lighting.We observe that for human eyes, besides individual images, the correlation (or the highlights variation) among input images also serves as an important hint to recognize * Contribute equally.

With the rapid development of deep learning, the research and application of micro-expression recognition are more and more extensive. In the existing solutions, the methods using the difference between onset frame and apex frame have high accuracy and low computational cost. But their extraction of dynamic features is not integrated with classification network, resulting in the lack of feedback from classification loss to the extracted dynamic features. In this paper, we propose a novel Displacement Generating Module (DGM) which uses convolution module to generate the displacement feature between onset frame and apex frame instead of traditional optical flow or dynamic image. The new DGM is integrated with the existing LEARNet to form an end-to-end micro-expression recognition network, where the classification loss can feed backward to the parameters of DGM to obtain better displacement features. We also present a random selection method of apex frame to increase the amount of training data, and present a normalization operation for the displacement features with different scales. Our new approach has been tested on SAMM, SMIC, CASME II datasets with LOSO evaluation method, and achieves 0.737 on UF1 and 0.726 on UAR, which is obviously higher than existing networks with optical flow and dynamic imaging techniques.

Features play an important role in the prediction tasks of e-commerce recommendations. To guarantee the consistency of off-line training and on-line serving, we usually utilize the same features that are both available. However, the consistency in turn neglects some discriminative features. For example, when estimating the conversion rate (CVR), i.e., the probability that a user would purchase the item if she clicked it, features like dwell time on the item detailed page are informative. However, CVR prediction should be conducted for on-line ranking before the click happens. Thus we cannot get such post-event features during serving.

Recommendation with multiple objectives is an important but difficult problem, where the coherent difficulty lies in the possible conflicts between objectives. In this case, multi-objective optimization is expected to be Pareto efficient, where no single objective can be further improved without hurting the others. However existing approaches to Pareto efficient multi-objective recommendation still lack good theoretical guarantees.

Ranking is a core task in recommender systems, which aims at providing an ordered list of items to users. Typically, a ranking function is learned from the labeled dataset to optimize the global performance, which produces a ranking score for each individual item. However, it may be sub-optimal because the scoring function applies to each item individually and does not explicitly consider the mutual influence between items, as well as the differences of users' preferences or intents. Therefore, we propose a personalized re-ranking model for recommender systems. The proposed re-ranking model can be easily deployed as a follow-up modular after any ranking algorithm, by directly using the existing ranking feature vectors. It directly optimizes the whole recommendation list by employing a transformer structure to efficiently encode the information of all items in the list. Specifically, the Transformer applies a self-attention mechanism that directly models the global relationships between any pair of items in the whole list. We confirm that the performance can be further improved by introducing pre-trained embedding to learn personalized encoding functions for different users. Experimental results on both offline benchmarks and real-world online e-commerce systems demonstrate the significant improvements of the proposed re-ranking model.

In this paper, we study the product title summarization problem in E-commerce applications for display on mobile devices. Comparing with conventional sentence summarization, product title summarization has some extra and essential constraints. For example, factual errors or loss of the key information are intolerable for E-commerce applications. Therefore, we abstract two more constraints for product title summarization: (i) do not introduce irrelevant information; (ii) retain the key information (e.g., brand name and commodity name). To address these issues, we propose a novel multi-source pointer network by adding a new knowledge encoder for pointer network. The first constraint is handled by pointer mechanism. For the second constraint, we restore the key information by copying words from the knowledge encoder with the help of the soft gating mechanism. For evaluation, we build a large collection of real-world product titles along with human-written short titles. Experimental results demonstrate that our model significantly outperforms the other baselines. Finally, online deployment of our proposed model has yielded a significant business impact, as measured by the click-through rate.