Bingkun Liu | Searcharxiv

arXiv2026

Refined Constraints on the Hard X-ray Polarization of the Crab Pulsar and Nebula Derived from an Extended XL-Calibur Dataset

We present updated hard X-ray polarization measurements of the Crab pulsar and nebula obtained with the balloon-borne polarimeter XL-Calibur in the ~19-64 keV energy range. During the flight, intermittent GPS-failure resulted in poorly constrained timing for ~38% of the Crab dataset. By implementing a new phase-recovery method that reconstructs timing during extended GPS-off intervals, phase tag data is recovered for ~95% of the GPS-off dataset, increasing the precision of the phase-resolved analysis. Phase-information for the data is recovered by using the Crab pulsar, with its 33 ms period, as an external timing source. Using a Markov-Chain Monte-Carlo framework to jointly fit phase offsets and frequency derivatives, sufficient phase accuracy is achieved, across multiple periods without GPS for a phase-resolved analysis. This enables inclusion of nearly the full dataset in the polarization study. The polarization degree of the nebular emission is found to be (27.7${\pm}$4.9)% at a polarization angle of 127.2{\deg}${\pm}$5.1{\deg} confirming previous XL-Calibur results and remaining aligned with the Crab's spin axis, consistent with synchrotron emission from the inner nebula. Phase-resolved measurements show that the off-pulse and bridge intervals exhibit a strong polarization, while the pulsar peaks, although weakly constrained, remain in agreement with the softer-energy trends of IXPE. These findings reinforce a scenario in which hard X-ray emission arises primarily in the nebular torus and wind regions. The successful recovery of precise phase tagging from GPS-off data demonstrates the capacity to use the pulsar as an external clock even in the case of sparsely populated data.

astro-ph.HE↗

arXiv2021

Few-Shot Model Adaptation for Customized Facial Landmark Detection, Segmentation, Stylization and Shadow Removal

Despite excellent progress has been made, the performance of deep learning based algorithms still heavily rely on specific datasets, which are difficult to extend due to labor-intensive labeling. Moreover, because of the advancement of new applications, initial definition of data annotations might not always meet the requirements of new functionalities. Thus, there is always a great demand in customized data annotations. To address the above issues, we propose the Few-Shot Model Adaptation (FSMA) framework and demonstrate its potential on several important tasks on Faces. The FSMA first acquires robust facial image embeddings by training an adversarial auto-encoder using large-scale unlabeled data. Then the model is equipped with feature adaptation and fusion layers, and adapts to the target task efficiently using a minimal amount of annotated images. The FSMA framework is prominent in its versatility across a wide range of facial image applications. The FSMA achieves state-of-the-art few-shot landmark detection performance and it offers satisfying solutions for few-shot face segmentation, stylization and facial shadow removal tasks for the first time.

cs.CV↗