arXiv subjects

Xuanyi Liu

Publications and source records attributed to Xuanyi Liu.

27 records · Page 2Linked to original sources

arXiv2024

X as Supervision: Contending with Depth Ambiguity in Unsupervised Monocular 3D Pose Estimation

Recent unsupervised methods for monocular 3D pose estimation have endeavored to reduce dependence on limited annotated 3D data, but most are solely formulated in 2D space, overlooking the inherent depth ambiguity issue. Due to the information loss in 3D-to-2D projection, multiple potential depths may exist, yet only some of them are plausible in human structure. To tackle depth ambiguity, we propose a novel unsupervised framework featuring a multi-hypothesis detector and multiple tailored pretext tasks. The detector extracts multiple hypotheses from a heatmap within a local window, effectively managing the multi-solution problem. Furthermore, the pretext tasks harness 3D human priors from the SMPL model to regularize the solution space of pose estimation, aligning it with the empirical distribution of 3D human structures. This regularization is partially achieved through a GCN-based discriminator within the discriminative learning, and is further complemented with synthetic images through rendering, ensuring plausible estimations. Consequently, our approach demonstrates state-of-the-art unsupervised 3D pose estimation performance on various human datasets. Further evaluations on data scale-up and one animal dataset highlight its generalization capabilities. Code will be available at https://github.com/Charrrrrlie/X-as-Supervision.

cs.CV↗

arXiv2024

Hybrid Mamba for Few-Shot Segmentation

Many few-shot segmentation (FSS) methods use cross attention to fuse support foreground (FG) into query features, regardless of the quadratic complexity. A recent advance Mamba can also well capture intra-sequence dependencies, yet the complexity is only linear. Hence, we aim to devise a cross (attention-like) Mamba to capture inter-sequence dependencies for FSS. A simple idea is to scan on support features to selectively compress them into the hidden state, which is then used as the initial hidden state to sequentially scan query features. Nevertheless, it suffers from (1) support forgetting issue: query features will also gradually be compressed when scanning on them, so the support features in hidden state keep reducing, and many query pixels cannot fuse sufficient support features; (2) intra-class gap issue: query FG is essentially more similar to itself rather than to support FG, i.e., query may prefer not to fuse support features but their own ones from the hidden state, yet the success of FSS relies on the effective use of support information. To tackle them, we design a hybrid Mamba network (HMNet), including (1) a support recapped Mamba to periodically recap the support features when scanning query, so the hidden state can always contain rich support information; (2) a query intercepted Mamba to forbid the mutual interactions among query pixels, and encourage them to fuse more support features from the hidden state. Consequently, the support information is better utilized, leading to better performance. Extensive experiments have been conducted on two public benchmarks, showing the superiority of HMNet. The code is available at https://github.com/Sam1224/HMNet.

cs.CV↗

arXiv2023

Proposal-Level Unsupervised Domain Adaptation for Open World Unbiased Detector

Open World Object Detection (OWOD) combines open-set object detection with incremental learning capabilities to handle the challenge of the open and dynamic visual world. Existing works assume that a foreground predictor trained on the seen categories can be directly transferred to identify the unseen categories' locations by selecting the top-k most confident foreground predictions. However, the assumption is hardly valid in practice. This is because the predictor is inevitably biased to the known categories, and fails under the shift in the appearance of the unseen categories. In this work, we aim to build an unbiased foreground predictor by re-formulating the task under Unsupervised Domain Adaptation, where the current biased predictor helps form the domains: the seen object locations and confident background locations as the source domain, and the rest ambiguous ones as the target domain. Then, we adopt the simple and effective self-training method to learn a predictor based on the domain-invariant foreground features, hence achieving unbiased prediction robust to the shift in appearance between the seen and unseen categories. Our approach's pipeline can adapt to various detection frameworks and UDA methods, empirically validated by OWOD evaluation, where we achieve state-of-the-art performance.

cs.CV↗

arXiv2023

All-polarization-maintaining linear cavity fiber lasers mode-locked by nonlinear polarization evolution in stretched pulse regime

Nonlinear polarization evolution (NPE) is among the most advanced techniques for obtaining ultrashort pulses with excellent optical performance. However, it is challenging to design environmentally stable NPE fiber oscillators using only polarization-maintaining (PM) fibers. Here, we use the same PM fiber and non-reciprocal phase shifter to design two different devices, which are capable of acting as effective NPE saturable absorbers (SAs) in two all-PM linear cavity fiber lasers. These two laser setups differ in the position of the non-reciprocal phase shifter, the presence of which is crucial for the proposed fiber lasers to reduce their mode-locking thresholds and achieve high repetition rates above 100 MHz. The mode-locking principle and pulse evolution in the laser cavity are investigated theoretically. The first all-PM fiber oscillator emits sub-200 fs stretched pulses with low peak powers. The second oscillator, with a simpler architecture, directly delivers stretched pulses with high peak powers, the spectral bandwidth greater than 30 nm, and the pulse duration less than 90 fs. To the best of our knowledge, 79 fs achieved in this design is the shortest pulse duration provided by PM fiber lasers using NPE mode-lockers.

physics.optics↗

arXiv2022

Robust mode-locking in a hybrid ultrafast laser based on nonlinear multimodal interference

We experimentally demonstrate the realization of a half-polarization-maintaining (half-PM) fiber laser, in which mode-locking is provided by a reflective multimode-interference saturable absorber (SA). In the specially designed SA, linearly polarized light is coupled into a 15-cm-long graded-index multimode fiber (GIMF) through the PM fiber, and then reflected back to the PM structure through a mirror pigtailed with a single-mode fiber (SMF). The modulation depth and saturation peak power are measured to be 1.5% and 0.6 W, respectively. The proposed SA device is incorporated into a novel half-PM erbium-doped fiber oscillator, which generates soliton pulses with 409 fs temporal duration at a 33.3 MHz repetition rate. The proposed fiber laser is compared with a conventional non-PM fiber laser mode-locked by nonlinear polarization evolution (NPE) in terms of optical properties such as spectral bandwidth, pulse duration, and stability performance. Short- and long-time stability tests and superior noise performance corroborate robust mode-locking in this setup.

physics.optics↗

arXiv2022

Observation of SQUID-like behavior in fiber laser with intra-cavity epsilon-near-zero effect

Establishing relations between fundamental effects in far-flung areas of physics is a subject of great interest in the current research. We here report realization of a novel photonic system akin to the radio-frequency superconducting quantum interference device (RF-SQUID), in a fiber laser cavity with epsilon-near-zero (ENZ) nanolayers as intra-cavity components. Emulating the RF-SQUID scheme, the photonic counterpart of the supercurrent, represented by the optical wave, circulates in the cavity, passing through effective optical potential barriers. Different ENZ wavelengths translate into distinct spectral outputs through the variation of cavity resonances, emulating the situation with a frequency-varying tank circuit in the RF-SQUID. Due to the presence of the ENZ element, the optical potential barrier is far lower for selected frequency components, granting them advantage in the gain-resource competition. The findings reported in this work provide a deeper insight into the ultrafast ENZ photonics, revealing a new path towards the design of nanophotonic on-chip devices with various operational functions, and offer a new approach to study superconducting and quantum-mechanical systems.

physics.optics↗

arXiv2022

Distance Guided Channel Weighting for Semantic Segmentation

Recent works have achieved great success in improving the performance of multiple computer vision tasks by capturing features with a high channel number utilizing deep neural networks. However, many channels of extracted features are not discriminative and contain a lot of redundant information. In this paper, we address above issue by introducing the Distance Guided Channel Weighting (DGCW) Module. The DGCW module is constructed in a pixel-wise context extraction manner, which enhances the discriminativeness of features by weighting different channels of each pixel's feature vector when modeling its relationship with other pixels. It can make full use of the high-discriminative information while ignore the low-discriminative information containing in feature maps, as well as capture the long-range dependencies. Furthermore, by incorporating the DGCW module with a baseline segmentation network, we propose the Distance Guided Channel Weighting Network (DGCWNet). We conduct extensive experiments to demonstrate the effectiveness of DGCWNet. In particular, it achieves 81.6% mIoU on Cityscapes with only fine annotated data for training, and also gains satisfactory performance on another two semantic segmentation datasets, i.e. Pascal Context and ADE20K. Code will be available soon at https://github.com/LanyunZhu/DGCWNet.

cs.CV↗

arXiv2021

A Robust and Novel Linear Fiber Laser Mode-locked by Nonlinear Polarization Evolution in All-polarization-maintaining Fibers

We demonstrate a novel, robust and compact fiber laser mode-locked by nonlinear polarization evolution (NPE) in polarization-maintaining (PM) fibers. The reflectivity of the artificial saturable absorber (SA) is analyzed to explain the mode-locking mechanism in the laser cavity. Experimentally, three linear laser schemes that feature repetition rates 94 MHz, 124 MHz and 133 MHz are systematically investigated. When the pump power is 1100 mW, the 124-MHz laser cavity delivers highly stable pulses with a single-pulse energy of 0.92 nJ. After the compression, the pulse duration obtained from the 124-MHz fiber laser is 250 fs, while the corresponding transform-limited pulse duration is 124 fs. The highest fundamental repetition rate that could be achieved in our experiment is 133 MHz, as mentioned above. The noise characterization has been performed with different cavity lengths and therefore different net-cavity dispersion. The 68-fs timing jitter and the 0.01% relative intensity noise (RIN) of the 133-MHz fiber laser have been realized integrated from 1 kHz to 10 MHz. Furthermore, the root-mean-square (RMS) power fluctuation is 0.35% in 2 hours, which implies superior stability of the output power. Thus, this linear fiber oscillator provides a competitive low-noise light source for optical applications appropriate for complex environments.

physics.optics↗

arXiv2021

Ultrafast Parallel LiDAR with Time-encoding and Spectral Scanning: Breaking the Time-of-flight Limit

Light detection and ranging (LiDAR) has been widely used in autonomous driving and large-scale manufacturing. Although state-of-the-art scanning LiDAR can perform long-range three-dimensional imaging, the frame rate is limited by both round-trip delay and the beam steering speed, hindering the development of high-speed autonomous vehicles. For hundred-meter level ranging applications, a several-time speedup is highly desirable. Here, we uniquely combine fiber-based encoders with wavelength-division multiplexing devices to implement all-optical time-encoding on the illumination light. Using this method, parallel detection and fast inertia-free spectral scanning can be achieved simultaneously with single-pixel detection. As a result, the frame rate of a scanning LiDAR can be multiplied with scalability. We demonstrate a 4.4-fold speedup for a maximum 75-m detection range, compared with a time-of-flight-limited laser ranging system. This approach has the potential to improve the velocity of LiDAR-based autonomous vehicles to the regime of hundred kilometers per hour and open up a new paradigm for ultrafast-frame-rate LiDAR imaging.

physics.optics↗