Search

In this work, we introduce the BBC-Oxford British Sign Language (BOBSL) dataset, a large-scale video collection of British Sign Language (BSL). BOBSL is an extended and publicly released dataset based on the BSL-1K dataset introduced in previous work. We describe the motivation for the dataset, together with statistics and available annotations. We conduct experiments to provide baselines for the tasks of sign recognition, sign language alignment, and sign language translation. Finally, we describe several strengths and limitations of the data from the perspectives of machine learning and linguistics, note sources of bias present in the dataset, and discuss potential applications of BOBSL in the context of sign language technology. The dataset is available at https://www.robots.ox.ac.uk/~vgg/data/bobsl/.

The objective of this work is to find temporal boundaries between signs in continuous sign language. Motivated by the paucity of annotation available for this task, we propose a simple yet effective algorithm to improve segmentation performance on unlabelled signing footage from a domain of interest. We make the following contributions: (1) We motivate and introduce the task of source-free domain adaptation for sign language segmentation, in which labelled source data is available for an initial training phase, but is not available during adaptation. (2) We propose the Changepoint-Modulated Pseudo-Labelling (CMPL) algorithm to leverage cues from abrupt changes in motion-sensitive feature space to improve pseudo-labelling quality for adaptation. (3) We showcase the effectiveness of our approach for category-agnostic sign segmentation, transferring from the BSLCORPUS to the BSL-1K and RWTH-PHOENIX-Weather 2014 datasets, where we outperform the prior state of the art., Comment: Appears in: 2021 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW'21). 11 pages

The recent rise of van der Waals (vdW) crystals has opened new prospects for studying versatile and exotic fundamental physics with future device applications such as twistronics. Even though the recent development on Angle-resolved photoemission spectroscopy (ARPES) with Nano-focusing optics, making clean surfaces and interfaces of chemically transferred crystals have been challenging to obtain high-resolution ARPES spectra. Here, we show that by employing nano-ARPES with submicron sized beam and polystyrene-assisted transfer followed by annealing process in ultra-high vacuum environment, remarkably clear ARPES spectral features such as spin-orbit splitting and band renormalization of CVD-grown, monolayered MoS2 can be measured. Our finding paves a way to exploit chemically transferred crystals for measuring high-resolution ARPES spectra to observe exotic quasi-particles in vdW heterostructures.

Recent progress in fine-grained gesture and action classification, and machine translation, point to the possibility of automated sign language recognition becoming a reality. A key stumbling block in making progress towards this goal is a lack of appropriate training data, stemming from the high complexity of sign annotation and a limited supply of qualified annotators. In this work, we introduce a new scalable approach to data collection for sign recognition in continuous videos. We make use of weakly-aligned subtitles for broadcast footage together with a keyword spotting method to automatically localise sign-instances for a vocabulary of 1,000 signs in 1,000 hours of video. We make the following contributions: (1) We show how to use mouthing cues from signers to obtain high-quality annotations from video data - the result is the BSL-1K dataset, a collection of British Sign Language (BSL) signs of unprecedented scale; (2) We show that we can use BSL-1K to train strong sign recognition models for co-articulated signs in BSL and that these models additionally form excellent pretraining for other sign languages and benchmarks - we exceed the state of the art on both the MSASL and WLASL benchmarks. Finally, (3) we propose new large-scale evaluation sets for the tasks of sign recognition and sign spotting and provide baselines which we hope will serve to stimulate research in this area., Comment: Appears in: European Conference on Computer Vision 2020 (ECCV 2020). 28 pages

Low resistivity (~3-24 mOhm.cm) with tunable n- and p-type phase pure Cu2O thin films have been grown by pulsed laser deposition at 25-200 0C by varying the background oxygen partial pressure (O2pp). Capacitance data obtained by electrochemical impedance spectroscopy was used to determine the conductivity (n- or p-type), carrier density, and flat band potentials for samples grown on indium tin oxide (ITO) at 25 0C. The Hall mobility of the n- and p-type Cu2O was estimated to be ~ 0.85 cm2.V-1s-1 and ~ 4.78 cm2.V-1s-1 respectively for samples grown on quartz substrate at 25 0C. An elevated substrate temperature ~ 200 0C with O2pp = 2 - 3 mTorr yielded p-type Cu2O films with six orders of magnitude higher resistivities in the range ~ 9 - 49 kOhm.cm and mobilities in the range ~ 13.5 - 22.2 cm2.V-1s-1. UV-Vis-NIR diffuse reflectance spectroscopy showed optical bandgaps of Cu2O films in the range of 1.76 to 2.15 eV depending on O2pp. Thin films grown at oxygen-rich conditions O2pp > 7 mTorr yielded mixed-phase copper oxide irrespective of the substrate temperatures and upon air annealing at 550 0C for 1 hour completely converted to CuO phase with n-type semiconducting properties (~12 Ohm.cm, ~1.50 cm2.V-1s-1). The as-grown p- and n-type Cu2O showed rectification and a photovoltaic (PV) response in solid junctions with n-ZnO and p-Si electrodes respectively. Our findings may create new opportunities for devising Cu2O based junctions requiring low process temperatures., Comment: 41 pages(including suppl. material), 7 figures (14 suppl. figures)

Copper-doped zinc sulphide (ZnS:Cu) thin films were synthesized through pulsed laser ablation in an inert background gas on stationary and rotating substrates, and a comprehensive opto-electrical characterisation is presented. The Cu$_x$Zn$_{1-x}$S films demonstrated comparable conductivity and transparency to other leading p-type transparent conducting materials, with a peak conductivity of 49.0 Scm$^{-1}$ and a hole mobility of 1.22 cm$^2$V$^{-1}$s$^{-1}$ for films alloyed with an x = 0.33 copper content. The most conducting films displayed a transparency of 71.8$\%$ over the visible range at a thickness of 100 nm, and band gaps were found in the range 3.22-3.52 eV, which showed a strong negative correlation with copper content. The effects of sulphur-rich rapid thermal annealing on the synthesized compound are reported, with films reliably displaying an increase in conductivity and carrier mobility. Films grown using a stationary substrate possessed large spatial thickness distributions and displayed sub-band gap absorption, which is discussed with respect to inhomogeneous copper substitution. Films deposited at 450$^\circ$C were found to be in the zincblende phase before and after annealing, with no occurrence of a phase change to wurtzite structure., Comment: 7 pages, 6 figures