科研成果 by Type: Conference Paper

Multispectral photometric stereo (MPS) aims at recovering the surface normal of a scene from a single-shot multispectral image captured under multispectral illuminations. Existing MPS methods adopt the Lambertian reflectance model to make the problem tractable, but it greatly limits their application to real-world surfaces. In this paper, we propose a deep neural network named NeuralMPS to solve the MPS problem under non-Lambertian spectral reflectances. Specifically, we present a spectral reflectance decomposition model to disentangle the spectral reflectance into a geometric component and a spectral component. With this decomposition, we show that the MPS problem for surfaces with a uniform material is equivalent to the conventional photometric stereo (CPS) with unknown light intensities. In this way, NeuralMPS reduces the difficulty of the non-Lambertian MPS problem by leveraging the well-studied non-Lambertian CPS methods. Experiments on both synthetic and real-world scenes demonstrate the effectiveness of our method.

As the popularity of hierarchical point forecast reconciliation methods increases, there is a growing interest in probabilistic forecast reconciliation. Many studies have utilized machine learning or deep learning techniques to implement probabilistic forecasting reconciliation and have made notable progress. However, these methods treat the reconciliation step as a fixed and hard post-processing step, leading to a trade-off between accuracy and coherency. In this paper, we propose a new approach for probabilistic forecast reconciliation. Unlike existing approaches, our proposed approach fuses the prediction step and reconciliation step into a deep learning framework, making the reconciliation step more flexible and soft by introducing the Kullback-Leibler divergence regularization term into the loss function. The approach is evaluated using three hierarchical time series datasets, which shows the advantages of our approach over other probabilistic forecast reconciliation methods.

The era was a form of "uninterrupted, irreversible, paratactic, cumulative, endless, and directional" time-reckoning concept (Kosmin 2018: 22) widely adopted in the Roman Greek East: there were provincial eras counting from the foundation of a province, as well as more customised epochs such as the 'Freiheitsära' of Amisos (Leschhorn 1993: 463-465), and the 'Kolonieära' of Sinope (Leschhorn 1993: 150-154). Curiously, Roman Sinope began with the colonial epoch of 45 BCE, but turned to a Lucullan epoch of 70 BCE in the Severan period (Kubitschek 1908: 67-72; Grant 1946: 12, 251; Leschhorn 1993: 151-162). Leschhorn suggested reasons for the switch, including internal rivalry and 'Gräzisierung' (Leschhorn 1993: 162). This paper explores the Graecisation hypothesis by first asking: why the Lucullan era? What was there to gain from the switch? Viewed from historical context, the Lucullan era happened to have been the time-reckoning method of Amastris, a well-developed regional hub in the second to third centuries CE (Marek 1993: 97-100; Brenier 2007), including contributions to several koina in Asia Minor. This paper suggests that Amastrian development may have influenced Sinopean institutions. Syncing time may be one way to maintain relations with a maritime hub with deep historical associations.  A second question this paper asks is how "Graecising" was the adoption of the Lucullan era. The Hellenistic amphora handles produced at Sinope stamped with the Seleucid era suggest that the use of this chronographic method predated the second century BCE (Saprykin & Fedoseev 1999: 135-143; Fedoseev 2019: 16-17). Instead of Caesarian colonists adopting a Hellenising time-reckoning, it may be that the Sinopean establishment modified (or revived) its epoch to mark both the new 'colonial' form of the Sinopean polity and the continuity of pre-colonial institutions (Magie 1950: 1267 n. 33; cf. Strab. 12.3.11 C54). In short, Sinope's changing eras may be viewed from an adaptive point of view.