Jiakai Zhang

Fuqiang Zhao, Yuheng Jiang, Kaixin Yao et al.

We have recently seen tremendous progress in the neural advances for photo-real human modeling and rendering. However, it's still challenging to integrate them into an existing mesh-based pipeline for downstream applications. In this paper, we present a comprehensive neural approach for high-quality reconstruction, compression, and rendering of human performances from dense multi-view videos. Our core intuition is to bridge the traditional animated mesh workflow with a new class of highly efficient neural techniques. We first introduce a neural surface reconstructor for high-quality surface generation in minutes. It marries the implicit volumetric rendering of the truncated signed distance field (TSDF) with multi-resolution hash encoding. We further propose a hybrid neural tracker to generate animated meshes, which combines explicit non-rigid tracking with implicit dynamic deformation in a self-supervised framework. The former provides the coarse warping back into the canonical space, while the latter implicit one further predicts the displacements using the 4D hash encoding as in our reconstructor. Then, we discuss the rendering schemes using the obtained animated meshes, ranging from dynamic texturing to lumigraph rendering under various bandwidth settings. To strike an intricate balance between quality and bandwidth, we propose a hierarchical solution by first rendering 6 virtual views covering the performer and then conducting occlusion-aware neural texture blending. We demonstrate the efficacy of our approach in a variety of mesh-based applications and photo-realistic free-view experiences on various platforms, i.e., inserting virtual human performances into real environments through mobile AR or immersively watching talent shows with VR headsets.

Xinhang Liu, Yan Zeng, Yifan Qin et al.

Cryo-electron microscopy (cryo-EM) allows for the high-resolution reconstruction of 3D structures of proteins and other biomolecules. Successful reconstruction of both shape and movement greatly helps understand the fundamental processes of life. However, it is still challenging to reconstruct the continuous motions of 3D structures from hundreds of thousands of noisy and randomly oriented 2D cryo-EM images. Recent advancements use Fourier domain coordinate-based neural networks to continuously model 3D conformations, yet they often struggle to capture local flexible regions accurately. We propose CryoFormer, a new approach for continuous heterogeneous cryo-EM reconstruction. Our approach leverages an implicit feature volume directly in the real domain as the 3D representation. We further introduce a novel query-based deformation transformer decoder to improve the reconstruction quality. Our approach is capable of refining pre-computed pose estimations and locating flexible regions. In experiments, our method outperforms current approaches on three public datasets (1 synthetic and 2 experimental) and a new synthetic dataset of PEDV spike protein. The code and new synthetic dataset will be released for better reproducibility of our results. Project page: https://cryoformer.github.io.

Yingjun Shen, Haizhao Dai, Qihe Chen et al.

Foundation models in computer vision have demonstrated exceptional performance in zero-shot and few-shot tasks by extracting multi-purpose features from large-scale datasets through self-supervised pre-training methods. However, these models often overlook the severe corruption in cryogenic electron microscopy (cryo-EM) images by high-level noises. We introduce DRACO, a Denoising-Reconstruction Autoencoder for CryO-EM, inspired by the Noise2Noise (N2N) approach. By processing cryo-EM movies into odd and even images and treating them as independent noisy observations, we apply a denoising-reconstruction hybrid training scheme. We mask both images to create denoising and reconstruction tasks. For DRACO's pre-training, the quality of the dataset is essential, we hence build a high-quality, diverse dataset from an uncurated public database, including over 270,000 movies or micrographs. After pre-training, DRACO naturally serves as a generalizable cryo-EM image denoiser and a foundation model for various cryo-EM downstream tasks. DRACO demonstrates the best performance in denoising, micrograph curation, and particle picking tasks compared to state-of-the-art baselines.

Jiakai Zhang, Qihe Chen, Yan Zeng et al.

In the past decade, deep conditional generative models have revolutionized the generation of realistic images, extending their application from entertainment to scientific domains. Single-particle cryo-electron microscopy (cryo-EM) is crucial in resolving near-atomic resolution 3D structures of proteins, such as the SARS- COV-2 spike protein. To achieve high-resolution reconstruction, a comprehensive data processing pipeline has been adopted. However, its performance is still limited as it lacks high-quality annotated datasets for training. To address this, we introduce physics-informed generative cryo-electron microscopy (CryoGEM), which for the first time integrates physics-based cryo-EM simulation with a generative unpaired noise translation to generate physically correct synthetic cryo-EM datasets with realistic noises. Initially, CryoGEM simulates the cryo-EM imaging process based on a virtual specimen. To generate realistic noises, we leverage an unpaired noise translation via contrastive learning with a novel mask-guided sampling scheme. Extensive experiments show that CryoGEM is capable of generating authentic cryo-EM images. The generated dataset can used as training data for particle picking and pose estimation models, eventually improving the reconstruction resolution.

Di Cao, Xianchen Wang, Junfeng Zhou et al.

Traditional Time Delay Neural Networks (TDNN) have achieved state-of-the-art performance at the cost of high computational complexity and slower inference speed, making them difficult to implement in an industrial environment. The Densely Connected Time Delay Neural Network (D-TDNN) with Context Aware Masking (CAM) module has proven to be an efficient structure to reduce complexity while maintaining system performance. In this paper, we propose a fast and lightweight model, LightCAM, which further adopts a depthwise separable convolution module (DSM) and uses multi-scale feature aggregation (MFA) for feature fusion at different levels. Extensive experiments are conducted on VoxCeleb dataset, the comparative results show that it has achieved an EER of 0.83 and MinDCF of 0.0891 in VoxCeleb1-O, which outperforms the other mainstream speaker verification methods. In addition, complexity analysis further demonstrates that the proposed architecture has lower computational cost and faster inference speed.

Liao Wang, Jiakai Zhang, Xinhang Liu et al.

Implicit neural representations such as Neural Radiance Field (NeRF) have focused mainly on modeling static objects captured under multi-view settings where real-time rendering can be achieved with smart data structures, e.g., PlenOctree. In this paper, we present a novel Fourier PlenOctree (FPO) technique to tackle efficient neural modeling and real-time rendering of dynamic scenes captured under the free-view video (FVV) setting. The key idea in our FPO is a novel combination of generalized NeRF, PlenOctree representation, volumetric fusion and Fourier transform. To accelerate FPO construction, we present a novel coarse-to-fine fusion scheme that leverages the generalizable NeRF technique to generate the tree via spatial blending. To tackle dynamic scenes, we tailor the implicit network to model the Fourier coefficients of timevarying density and color attributes. Finally, we construct the FPO and train the Fourier coefficients directly on the leaves of a union PlenOctree structure of the dynamic sequence. We show that the resulting FPO enables compact memory overload to handle dynamic objects and supports efficient fine-tuning. Extensive experiments show that the proposed method is 3000 times faster than the original NeRF and achieves over an order of magnitude acceleration over SOTA while preserving high visual quality for the free-viewpoint rendering of unseen dynamic scenes.

Jiakai Zhang, Liao Wang, Xinhang Liu et al.

Some of the most exciting experiences that Metaverse promises to offer, for instance, live interactions with virtual characters in virtual environments, require real-time photo-realistic rendering. 3D reconstruction approaches to rendering, active or passive, still require extensive cleanup work to fix the meshes or point clouds. In this paper, we present a neural volumography technique called neural volumetric video or NeuVV to support immersive, interactive, and spatial-temporal rendering of volumetric video contents with photo-realism and in real-time. The core of NeuVV is to efficiently encode a dynamic neural radiance field (NeRF) into renderable and editable primitives. We introduce two types of factorization schemes: a hyper-spherical harmonics (HH) decomposition for modeling smooth color variations over space and time and a learnable basis representation for modeling abrupt density and color changes caused by motion. NeuVV factorization can be integrated into a Video Octree (VOctree) analogous to PlenOctree to significantly accelerate training while reducing memory overhead. Real-time NeuVV rendering further enables a class of immersive content editing tools. Specifically, NeuVV treats each VOctree as a primitive and implements volume-based depth ordering and alpha blending to realize spatial-temporal compositions for content re-purposing. For example, we demonstrate positioning varied manifestations of the same performance at different 3D locations with different timing, adjusting color/texture of the performer's clothing, casting spotlight shadows and synthesizing distance falloff lighting, etc, all at an interactive speed. We further develop a hybrid neural-rasterization rendering framework to support consumer-level VR headsets so that the aforementioned volumetric video viewing and editing, for the first time, can be conducted immersively in virtual 3D space.

Fuqiang Zhao, Wei Yang, Jiakai Zhang et al.

Recent neural human representations can produce high-quality multi-view rendering but require using dense multi-view inputs and costly training. They are hence largely limited to static models as training each frame is infeasible. We present HumanNeRF - a generalizable neural representation - for high-fidelity free-view synthesis of dynamic humans. Analogous to how IBRNet assists NeRF by avoiding per-scene training, HumanNeRF employs an aggregated pixel-alignment feature across multi-view inputs along with a pose embedded non-rigid deformation field for tackling dynamic motions. The raw HumanNeRF can already produce reasonable rendering on sparse video inputs of unseen subjects and camera settings. To further improve the rendering quality, we augment our solution with an appearance blending module for combining the benefits of both neural volumetric rendering and neural texture blending. Extensive experiments on various multi-view dynamic human datasets demonstrate the generalizability and effectiveness of our approach in synthesizing photo-realistic free-view humans under challenging motions and with very sparse camera view inputs.

Jiakai Zhang, Xinhang Liu, Xinyi Ye et al.

Generating free-viewpoint videos is critical for immersive VR/AR experience but recent neural advances still lack the editing ability to manipulate the visual perception for large dynamic scenes. To fill this gap, in this paper we propose the first approach for editable photo-realistic free-viewpoint video generation for large-scale dynamic scenes using only sparse 16 cameras. The core of our approach is a new layered neural representation, where each dynamic entity including the environment itself is formulated into a space-time coherent neural layered radiance representation called ST-NeRF. Such layered representation supports fully perception and realistic manipulation of the dynamic scene whilst still supporting a free viewing experience in a wide range. In our ST-NeRF, the dynamic entity/layer is represented as continuous functions, which achieves the disentanglement of location, deformation as well as the appearance of the dynamic entity in a continuous and self-supervised manner. We propose a scene parsing 4D label map tracking to disentangle the spatial information explicitly, and a continuous deform module to disentangle the temporal motion implicitly. An object-aware volume rendering scheme is further introduced for the re-assembling of all the neural layers. We adopt a novel layered loss and motion-aware ray sampling strategy to enable efficient training for a large dynamic scene with multiple performers, Our framework further enables a variety of editing functions, i.e., manipulating the scale and location, duplicating or retiming individual neural layers to create numerous visual effects while preserving high realism. Extensive experiments demonstrate the effectiveness of our approach to achieve high-quality, photo-realistic, and editable free-viewpoint video generation for dynamic scenes.

Quan Meng, Jiakai Zhang, Qiang Hu et al.

We present a novel real-time line segment detection scheme called Line Graph Neural Network (LGNN). Existing approaches require a computationally expensive verification or postprocessing step. Our LGNN employs a deep convolutional neural network (DCNN) for proposing line segment directly, with a graph neural network (GNN) module for reasoning their connectivities. Specifically, LGNN exploits a new quadruplet representation for each line segment where the GNN module takes the predicted candidates as vertexes and constructs a sparse graph to enforce structural context. Compared with the state-of-the-art, LGNN achieves near real-time performance without compromising accuracy. LGNN further enables time-sensitive 3D applications. When a 3D point cloud is accessible, we present a multi-modal line segment classification technique for extracting a 3D wireframe of the environment robustly and efficiently.

Yanghua Jin, Jiakai Zhang, Minjun Li et al.

Automatic generation of facial images has been well studied after the Generative Adversarial Network (GAN) came out. There exists some attempts applying the GAN model to the problem of generating facial images of anime characters, but none of the existing work gives a promising result. In this work, we explore the training of GAN models specialized on an anime facial image dataset. We address the issue from both the data and the model aspect, by collecting a more clean, well-suited dataset and leverage proper, empirical application of DRAGAN. With quantitative analysis and case studies we demonstrate that our efforts lead to a stable and high-quality model. Moreover, to assist people with anime character design, we build a website (http://make.girls.moe) with our pre-trained model available online, which makes the model easily accessible to general public.

Rahul Mitra, Jiakai Zhang, Sanath Narayan et al.

We propose a convolutional neural network (ConvNet) based approach for learning local image descriptors which can be used for significantly improved patch matching and 3D reconstructions. A multi-resolution ConvNet is used for learning keypoint descriptors. We also propose a new dataset consisting of an order of magnitude more number of scenes, images, and positive and negative correspondences compared to the currently available Multi-View Stereo (MVS) [18] dataset. The new dataset also has better coverage of the overall viewpoint, scale, and lighting changes in comparison to the MVS dataset. We evaluate our approach on publicly available datasets, such as Oxford Affine Covariant Regions Dataset (ACRD) [12], MVS [18], Synthetic [6] and Strecha [15] datasets to quantify the image descriptor performance. Scenes from the Oxford ACRD, MVS and Synthetic datasets are used for evaluating the patch matching performance of the learnt descriptors while the Strecha dataset is used to evaluate the 3D reconstruction task. Experiments show that the proposed descriptor outperforms the current state-of-the-art descriptors in both the evaluation tasks.

Jiakai Zhang, Kyunghyun Cho

One way to approach end-to-end autonomous driving is to learn a policy function that maps from a sensory input, such as an image frame from a front-facing camera, to a driving action, by imitating an expert driver, or a reference policy. This can be done by supervised learning, where a policy function is tuned to minimize the difference between the predicted and ground-truth actions. A policy function trained in this way however is known to suffer from unexpected behaviours due to the mismatch between the states reachable by the reference policy and trained policy functions. More advanced algorithms for imitation learning, such as DAgger, addresses this issue by iteratively collecting training examples from both reference and trained policies. These algorithms often requires a large number of queries to a reference policy, which is undesirable as the reference policy is often expensive. In this paper, we propose an extension of the DAgger, called SafeDAgger, that is query-efficient and more suitable for end-to-end autonomous driving. We evaluate the proposed SafeDAgger in a car racing simulator and show that it indeed requires less queries to a reference policy. We observe a significant speed up in convergence, which we conjecture to be due to the effect of automated curriculum learning.

Mariusz Bojarski, Davide Del Testa, Daniel Dworakowski et al.

We trained a convolutional neural network (CNN) to map raw pixels from a single front-facing camera directly to steering commands. This end-to-end approach proved surprisingly powerful. With minimum training data from humans the system learns to drive in traffic on local roads with or without lane markings and on highways. It also operates in areas with unclear visual guidance such as in parking lots and on unpaved roads. The system automatically learns internal representations of the necessary processing steps such as detecting useful road features with only the human steering angle as the training signal. We never explicitly trained it to detect, for example, the outline of roads. Compared to explicit decomposition of the problem, such as lane marking detection, path planning, and control, our end-to-end system optimizes all processing steps simultaneously. We argue that this will eventually lead to better performance and smaller systems. Better performance will result because the internal components self-optimize to maximize overall system performance, instead of optimizing human-selected intermediate criteria, e.g., lane detection. Such criteria understandably are selected for ease of human interpretation which doesn't automatically guarantee maximum system performance. Smaller networks are possible because the system learns to solve the problem with the minimal number of processing steps. We used an NVIDIA DevBox and Torch 7 for training and an NVIDIA DRIVE(TM) PX self-driving car computer also running Torch 7 for determining where to drive. The system operates at 30 frames per second (FPS).