We study the task of image inpainting, which is to fill in the missing region of an incomplete image with plausible contents. To this end, we propose a learning-based approach to generate visually coherent completion given a high-resolution image with missing components. In order to overcome the difficulty to directly learn the distribution of highdimensional image data, we divide the task into inference and translation as two separate steps and model each step with a deep neural network. We also use simple heuristics to guide the propagation of local textures from the boundary to the hole. We show that, by using such techniques, inpainting reduces to the problem of learning two image-feature translation functions in much smaller space and hence easier to train. We evaluate our method on several public datasets and show that we generate results of better visual quality than previous state-of-the-art methods.* indicates equal contribution arXiv:1711.08590v5 [cs.CV]

We propose a method for unsupervised video object segmentation by transferring the knowledge encapsulated in image-based instance embedding networks. The instance embedding network produces an embedding vector for each pixel that enables identifying all pixels belonging to the same object. Though trained on static images, the instance embeddings are stable over consecutive video frames, which allows us to link objects together over time. Thus, we adapt the instance networks trained on static images to video object segmentation and incorporate the embeddings with objectness and optical flow features, without model retraining or online fine-tuning. The proposed method outperforms state-of-the-art unsupervised segmentation methods in the DAVIS dataset and the FBMS dataset.

Using tetrathienoacene (TT) and diketopyrrolopyrrole (DPP) units as donor and acceptor blocks, three new low band gap (LBG) conjugated copolymers (PTTDPO, PTTDPS and PTTDPSe) with different heterocycle bridges (furan, thiophene and selenophene) were designed and synthesized by Stille coupling polymerization reactions. Their structures were verified by 1 HNMR spectroscopy and elemental analysis.All these copolymers exhibit broad absorption bands with small band gaps. UV-vis absorption spectra and cyclic voltammetry (CV) measurements indicated that the selenophene inclusion resulted in a reduction in the band gap, which could be attributed to a reduction of oxidation potentials and an increase of the electron affinities of the related copolymer. The density functional theory (DFT) calculations showed that PTTDPSe with selenophene bridges favoured a much more planar conformation than PTTDPO and PTTDPS, which afforded a higher hole mobility of 7.9 Â 10 À4 cm 2 V À1 s À1 . Photovoltaic properties of the copolymers blended with [6,6]-phenyl-C 71 -butyric acid methyl ester (PC 71 BM) as an electron acceptor were investigated. The polymer solar cell (PSC) based on the structure of ITO/PEDOT:PSS/PTTDPSe:PC 71 BM (1 : 1.5, w/w)/Ca/Al exhibited a high power conversion efficiency (PCE) of 5.68% with an improved short circuit current density (J sc ) of 15.62 mA cm À2 . The primary resultsshow that changing heterocycle bridges with different electron-donating ability can easily and finely tune the optical absorptions, band gaps and energy levels of DPP-containing copolymers. The results also demonstrate that the TT unit is a new and promising electron-donating donor block for constructing highly efficient LBG photovoltaic materials.

Replacing benzodithiophene (BDT) with a naphthodithiophene (NDT) building block, PNDTDPP exhibited enhanced photovoltaic performance with a PCE of 5.37% when compared with BDT-based copolymer PBDDPP, which gave a PCE of 2.91% in conventional device structures. Encouragingly, the obtained inverted PSC with PBDDPP achieved an impressively high PCE of 6.92%.