DynaSurfGS: Dynamic Surface Reconstruction with Planar-based Gaussian Splatting

Dynamic scene reconstruction has garnered significant attention in recent years due to its capabilities in high-quality and real-time rendering. Among various methodologies, constructing a 4D spatial-temporal representation, such as 4D-GS, has gained popularity for its high-quality rendered images. However, these methods often produce suboptimal surfaces, as the discrete 3D Gaussian point clouds fail to align with the object's surface precisely. To address this problem, we propose DynaSurfGS to achieve both photorealistic rendering and high-fidelity surface reconstruction of dynamic scenarios. Specifically, the DynaSurfGS framework first incorporates Gaussian features from 4D neural voxels with the planar-based Gaussian Splatting to facilitate precise surface reconstruction. It leverages normal regularization to enforce the smoothness of the surface of dynamic objects. It also incorporates the as-rigid-as-possible (ARAP) constraint to maintain the approximate rigidity of local neighborhoods of 3D Gaussians between timesteps and ensure that adjacent 3D Gaussians remain closely aligned throughout. Extensive experiments demonstrate that DynaSurfGS surpasses state-of-the-art methods in both high-fidelity surface reconstruction and photorealistic rendering.

GraftNet: An Engineering Implementation of CNN for Fine-grained Multi-label Task

Multi-label networks with branches are proved to perform well in both accuracy and speed, but lacks flexibility in providing dynamic extension onto new labels due to the low efficiency of re-work on annotating and training. For multi-label classification task, to cover new labels we need to annotate not only newly collected images, but also the previous whole dataset to check presence of these new labels. Also training on whole re-annotated dataset costs much time. In order to recognize new labels more effectively and accurately, we propose GraftNet, which is a customizable tree-like network with its trunk pretrained with a dynamic graph for generic feature extraction, and branches separately trained on sub-datasets with single label to improve accuracy. GraftNet could reduce cost, increase flexibility, and incrementally handle new labels. Experimental results show that it has good performance on our human attributes recognition task, which is fine-grained multi-label classification.

Diversity-Generated Image Inpainting with Style Extraction

The latest methods based on deep learning have achieved amazing results regarding the complex work of inpainting large missing areas in an image. This type of method generally attempts to generate one single "optimal" inpainting result, ignoring many other plausible results. However, considering the uncertainty of the inpainting task, one sole result can hardly be regarded as a desired regeneration of the missing area. In view of this weakness, which is related to the design of the previous algorithms, we propose a novel deep generative model equipped with a brand new style extractor which can extract the style noise (a latent vector) from the ground truth image. Once obtained, the extracted style noise and the ground truth image are both input into the generator. We also craft a consistency loss that guides the generated image to approximate the ground truth. Meanwhile, the same extractor captures the style noise from the generated image, which is forced to approach the input noise according to the consistency loss. After iterations, our generator is able to learn the styles corresponding to multiple sets of noise. The proposed model can generate a (sufficiently large) number of inpainting results consistent with the context semantics of the image. Moreover, we check the effectiveness of our model on three databases, i.e., CelebA, Agricultural Disease, and MauFlex. Compared to state-of-the-art inpainting methods, this model is able to offer desirable inpainting results with both a better quality and higher diversity, for example, on the human face, we can even output the different gaze angles of the eyes and whether they have glasses, etc. The code and model will be made available on https://github.com/vivitsai/SEGAN.