In this paper, we address the detection of co-occurring salient objects (CoSOD) in an image group using frequency statistics in an unsupervised manner, which further enable us to develop a semi-supervised method. While previous works have mostly focused on fully supervised CoSOD, less attention has been allocated to detecting co-salient objects when limited segmentation annotations are available for training. Our simple yet effective unsupervised method US-CoSOD combines the object co-occurrence frequency statistics of unsupervised single-image semantic segmentations with salient foreground detections using self-supervised feature learning. For the first time, we show that a large unlabeled dataset e.g. ImageNet-1k can be effectively leveraged to significantly improve unsupervised CoSOD performance. Our unsupervised model is a great pre-training initialization for our semi-supervised model SS-CoSOD, especially when very limited labeled data is available for training. To avoid propagating erroneous signals from predictions on unlabeled data, we propose a confidence estimation module to guide our semi-supervised training. Extensive experiments on three CoSOD benchmark datasets show that both of our unsupervised and semi-supervised models outperform the corresponding state-of-the-art models by a significant margin (e.g., on the Cosal2015 dataset, our US-CoSOD model has an 8.8% F-measure gain over a SOTA unsupervised co-segmentation model and our SS-CoSOD model has an 11.81% F-measure gain over a SOTA semi-supervised CoSOD model).

We present a model for predicting visual attention during the free viewing of graphic design documents. While existing works on this topic have aimed at predicting static saliency of graphic designs, our work is the first attempt to predict both spatial attention and dynamic temporal order in which the document regions are fixated by gaze using a deep learning based model. We propose a two-stage model for predicting dynamic attention on such documents, with webpages being our primary choice of document design for demonstration. In the first stage, we predict the saliency maps for each of the document components (e.g. logos, banners, texts, etc. for webpages) conditioned on the type of document layout. These component saliency maps are then jointly used to predict the overall document saliency. In the second stage, we use these layout-specific component saliency maps as the state representation for an inverse reinforcement learning model of fixation scanpath prediction during document viewing. To test our model, we collected a new dataset consisting of eye movements from 41 people freely viewing 450 webpages (the largest dataset of its kind). Experimental results show that our model outperforms existing models in both saliency and scanpath prediction for webpages, and also generalizes very well to other graphic design documents such as comics, posters, mobile UIs, etc. and natural images.

The explosive growth of wireless data and traffic, accompanied by the rapid advancements in intelligence and processing power of user equipments (UEs), has paved the way for device-to-device (D2D) communication technology to surface as a promising solution. One major benefit is that users can collaboratively cache and share content to reduce costs associated with backhaul links. In this paper, we explore different approaches to cache content for users in a D2D enabled environment and propose a novel two-stacked approach to achieve a higher cache-hit ratio while leveraging advancements in deep learning. We propose the ‘R2-D2D’ framework, wherein we use Long Short-Term Memory (LSTM) networks stacked with a recently developed omni-scale convolutional neural network (CNN) for making the caching decision. Unlike most previous works, the proposed system model works without any apriori knowledge like file popularity distribution, or any assumptions like stationarity of the environment. Our experiments show that the proposed framework learns well from historical information, obtaining an overall average D2D cache hit ratio of 0.418 when 5000 timesteps of historical information were provided, outperforming a recently proposed neural network collaborative filtering (NCF) framework by approximately 10% to 25%.

In this study, the authors aim to colourise a greyscale image using a fully automated framework which retrieves similar images from a reference database and then transfers the colour from the most similar retrieved images to perform colourisation. Inspired by the recent success of deep learning techniques in extracting semantic information from images, they first use fc7 features from AlexNet to retrieve similar images from the reference database. Top‐ k retrieved images are considered for colour transfer to the target greyscale image, using various pixel level features. The images which result from the previous step are given a colour enhancement with Reinhard stain normalisation. They follow a pixel‐wise colour saturation based averaging technique to impart colour at pixel level. The final image is rectified using joint bilateral filtering. The resulting coloured images have a realistic appearance, similar in quality to the original coloured images. The proposed method outperforms several previous colourisation techniques, yielding superior performance both quantitatively and qualitatively. The method also enhances low‐contrast images.