This paper proposes an end-to-end framework for generating 3D human pose datasets using Neural Radiance Fields (NeRF). Public datasets generally have limited diversity in terms of human poses and camera viewpoints, largely due to the resource-intensive nature of collecting 3D human pose data. As a result, pose estimators trained on public datasets significantly underperform when applied to unseen out-of-distribution samples. Previous works proposed augmenting public datasets by generating 2D-3D pose pairs or rendering a large amount of random data. Such approaches either overlook image rendering or result in suboptimal datasets for pre-trained models. Here we propose PoseGen, which learns to generate a dataset (human 3D poses and images) with a feedback loss from a given pre-trained pose estimator. In contrast to prior art, our generated data is optimized to improve the robustness of the pre-trained model. The objective of PoseGen is to learn a distribution of data that maximizes the prediction error of a given pre-trained model. As the learned data distribution contains OOD samples of the pre-trained model, sampling data from such a distribution for further fine-tuning a pre-trained model improves the generalizability of the model. This is the first work that proposes NeRFs for 3D human data generation. NeRFs are data-driven and do not require 3D scans of humans. Therefore, using NeRF for data generation is a new direction for convenient user-specific data generation. Our extensive experiments show that the proposed PoseGen improves two baseline models (SPIN and HybrIK) on four datasets with an average 6% relative improvement.
In this paper, we propose a novel time of arrival (TOA) estimator for multiple-input-multiple-output (MIMO) backscatter channels in closed form. The proposed estimator refines the estimation precision from the topological structure of the MIMO backscatter channels, and can considerably enhance the estimation accuracy. Particularly, we show that for the general $M \times N$ bistatic topology, the mean square error (MSE) is $\frac{M+N-1}{MN}\sigma^2_0$, and for the general $M \times M$ monostatic topology, it is $\frac{2M-1}{M^2}\sigma^2_0$ for the diagonal subchannels, and $\frac{M-1}{M^2}\sigma^2_0$ for the off-diagonal subchannels, where $\sigma^2_0$ is the MSE of the conventional least square estimator. In addition, we derive the Cramer-Rao lower bound (CRLB) for MIMO backscatter TOA estimation which indicates that the proposed estimator is optimal. Simulation results verify that the proposed TOA estimator can considerably improve both estimation and positioning accuracy, especially when the MIMO scale is large.
Facial action units (FAUs) are critical for fine-grained facial expression analysis. Although FAU detection has been actively studied using ideally high quality images, it was not thoroughly studied under heavily occluded conditions. In this paper, we propose the first occlusion-robust FAU recognition method to maintain FAU detection performance under heavy occlusions. Our novel approach takes advantage of rich information from the latent space of masked autoencoder (MAE) and transforms it into FAU features. Bypassing the occlusion reconstruction step, our model efficiently extracts FAU features of occluded faces by mining the latent space of a pretrained masked autoencoder. Both node and edge-level knowledge distillation are also employed to guide our model to find a mapping between latent space vectors and FAU features. Facial occlusion conditions, including random small patches and large blocks, are thoroughly studied. Experimental results on BP4D and DISFA datasets show that our method can achieve state-of-the-art performances under the studied facial occlusion, significantly outperforming existing baseline methods. In particular, even under heavy occlusion, the proposed method can achieve comparable performance as state-of-the-art methods under normal conditions.
Modern autonomous vehicles rely heavily on mechanical LiDARs for perception. Current perception methods generally require 360{\deg} point clouds, collected sequentially as the LiDAR scans the azimuth and acquires consecutive wedge-shaped slices. The acquisition latency of a full scan (~ 100ms) may lead to outdated perception which is detrimental to safe operation. Recent streaming perception works proposed directly processing LiDAR slices and compensating for the narrow field of view (FOV) of a slice by reusing features from preceding slices. These works, however, are all based on a single modality and require past information which may be outdated. Meanwhile, images from high-frequency cameras can support streaming models as they provide a larger FoV compared to a LiDAR slice. However, this difference in FoV complicates sensor fusion. To address this research gap, we propose an innovative camera-LiDAR streaming 3D object detection framework that uses camera images instead of past LiDAR slices to provide an up-to-date, dense, and wide context for streaming perception. The proposed method outperforms prior streaming models on the challenging NuScenes benchmark. It also outperforms powerful full-scan detectors while being much faster. Our method is shown to be robust to missing camera images, narrow LiDAR slices, and small camera-LiDAR miscalibration.
Outdoor-to-indoor communications in millimeter-wave (mmWave) cellular networks have been one challenging research problem due to the severe attenuation and the high penetration loss caused by the propagation characteristics of mmWave signals. We propose a viable solution to implement the outdoor-to-indoor mmWave communication system with the aid of an active intelligent transmitting surface (active-ITS), where the active-ITS allows the incoming signal from an outdoor base station (BS) to pass through the surface and be received by the indoor user-equipments (UEs) after shifting its phase and magnifying its amplitude. Then, the problem of joint precoding of the BS and active-ITS is investigated to maximize the weighted sum-rate (WSR) of the communication system. An efficient block coordinate descent (BCD) based algorithm is developed to solve it with the suboptimal solutions in nearly closed-forms. In addition, to reduce the size and hardware cost of an active-ITS, we provide a block-amplifying architecture to partially remove the circuit components for power-amplifying, where multiple transmissive-type elements (TEs) in each block share a same power amplifier. Simulations indicate that active-ITS has the potential of achieving a given performance with much fewer TEs compared to the passive-ITS under the same total system power consumption, which makes it suitable for application to the size-limited and aesthetic-needed scenario, and the inevitable performance degradation caused by the block-amplifying architecture is acceptable.
We propose a Transformer-based NeRF (TransNeRF) to learn a generic neural radiance field conditioned on observed-view images for the novel view synthesis task. By contrast, existing MLP-based NeRFs are not able to directly receive observed views with an arbitrary number and require an auxiliary pooling-based operation to fuse source-view information, resulting in the missing of complicated relationships between source views and the target rendering view. Furthermore, current approaches process each 3D point individually and ignore the local consistency of a radiance field scene representation. These limitations potentially can reduce their performance in challenging real-world applications where large differences between source views and a novel rendering view may exist. To address these challenges, our TransNeRF utilizes the attention mechanism to naturally decode deep associations of an arbitrary number of source views into a coordinate-based scene representation. Local consistency of shape and appearance are considered in the ray-cast space and the surrounding-view space within a unified Transformer network. Experiments demonstrate that our TransNeRF, trained on a wide variety of scenes, can achieve better performance in comparison to state-of-the-art image-based neural rendering methods in both scene-agnostic and per-scene finetuning scenarios especially when there is a considerable gap between source views and a rendering view.
Skin cancer is one of the most common types of malignancy, affecting a large population and causing a heavy economic burden worldwide. Over the last few years, computer-aided diagnosis has been rapidly developed and make great progress in healthcare and medical practices due to the advances in artificial intelligence. However, most studies in skin cancer detection keep pursuing high prediction accuracies without considering the limitation of computing resources on portable devices. In this case, knowledge distillation (KD) has been proven as an efficient tool to help improve the adaptability of lightweight models under limited resources, meanwhile keeping a high-level representation capability. To bridge the gap, this study specifically proposes a novel method, termed SSD-KD, that unifies diverse knowledge into a generic KD framework for skin diseases classification. Our method models an intra-instance relational feature representation and integrates it with existing KD research. A dual relational knowledge distillation architecture is self-supervisedly trained while the weighted softened outputs are also exploited to enable the student model to capture richer knowledge from the teacher model. To demonstrate the effectiveness of our method, we conduct experiments on ISIC 2019, a large-scale open-accessed benchmark of skin diseases dermoscopic images. Experiments show that our distilled lightweight model can achieve an accuracy as high as 85% for the classification tasks of 8 different skin diseases with minimal parameters and computing requirements. Ablation studies confirm the effectiveness of our intra- and inter-instance relational knowledge integration strategy. Compared with state-of-the-art knowledge distillation techniques, the proposed method demonstrates improved performances for multi-diseases classification on the large-scale dermoscopy database.
Recent advances in wearable devices and Internet-of-Things (IoT) have led to massive growth in sensor data generated in edge devices. Labeling such massive data for classification tasks has proven to be challenging. In addition, data generated by different users bear various personal attributes and edge heterogeneity, rendering it impractical to develop a global model that adapts well to all users. Concerns over data privacy and communication costs also prohibit centralized data accumulation and training. This paper proposes a novel personalized semi-supervised federated learning (SemiPFL) framework to support edge users having no label or limited labeled datasets and a sizable amount of unlabeled data that is insufficient to train a well-performing model. In this work, edge users collaborate to train a hyper-network in the server, generating personalized autoencoders for each user. After receiving updates from edge users, the server produces a set of base models for each user, which the users locally aggregate them using their own labeled dataset. We comprehensively evaluate our proposed framework on various public datasets and demonstrate that SemiPFL outperforms state-of-art federated learning frameworks under the same assumptions. We also show that the solution performs well for users without labeled datasets or having limited labeled datasets and increasing performance for increased labeled data and number of users, signifying the effectiveness of SemiPFL for handling edge heterogeneity and limited annotation. By leveraging personalized semi-supervised learning, SemiPFL dramatically reduces the need for annotating data and preserving privacy in a wide range of application scenarios, from wearable health to IoT.
Intelligent reflecting surfaces (IRSs) have emerged as a promising economical solution to implement cell-free networks. However, the performance gains achieved by IRSs critically depend on smartly tuned passive beamforming based on the assumption that the accurate channel state information (CSI) knowledge is available at the central processing unit (CPU), which is practically impossible. Thus, in this paper, we investigate the impact of the CSI uncertainty on IRS-assisted cell-free networks. We adopt a stochastic method to cope with the CSI uncertainty by maximizing the expectation of the sum-rate, which guarantees the robust performance over the average. Accordingly, an average sum-rate maximization problem is formulated, which is non-convex and arduous to obtain its optimal solution due to the coupled variables and the expectation operation with respect to CSI uncertainties. As a compromising approach, we develop an efficient robust joint design algorithm with low-complexity. Particularly, the original problem is equivalently transformed into a tractable form by employing algebraic manipulations. Then, the locally optimal solution can be obtained iteratively. We further prove that the CSI uncertainty has no direct impact on the optimizing of the passive reflecting beamforming. It is worth noting that the investigated scenario is flexible and general in communications, thus the proposed algorithm can act as a general framework to solve various sum-rate maximization problems. Simulation results demonstrate that IRSs can achieve considerable data rate improvement for conventional cell-free networks, and confirm the resilience of the proposed algorithm against the CSI uncertainty.
This paper addresses the problem of cross-dataset generalization of 3D human pose estimation models. Testing a pre-trained 3D pose estimator on a new dataset results in a major performance drop. Previous methods have mainly addressed this problem by improving the diversity of the training data. We argue that diversity alone is not sufficient and that the characteristics of the training data need to be adapted to those of the new dataset such as camera viewpoint, position, human actions, and body size. To this end, we propose AdaptPose, an end-to-end framework that generates synthetic 3D human motions from a source dataset and uses them to fine-tune a 3D pose estimator. AdaptPose follows an adversarial training scheme. From a source 3D pose the generator generates a sequence of 3D poses and a camera orientation that is used to project the generated poses to a novel view. Without any 3D labels or camera information AdaptPose successfully learns to create synthetic 3D poses from the target dataset while only being trained on 2D poses. In experiments on the Human3.6M, MPI-INF-3DHP, 3DPW, and Ski-Pose datasets our method outperforms previous work in cross-dataset evaluations by 14% and previous semi-supervised learning methods that use partial 3D annotations by 16%.