Abstract

We address the visual relocalization problem of predicting the location and camera orientation or pose (6DOF) of the given input scene. We propose a method based on how humans determine their location using the visible landmarks. We define anchor points uniformly across the route map and propose a deep learning architecture which predicts the most relevant anchor point present in the scene as well as the relative offsets with respect to it. The relevant anchor point need not be the nearest anchor point to the ground truth location, as it might not be visible due to the pose. Hence we propose a multi task loss function, which discovers the relevant anchor point, without needing the ground truth for it. We validate the effectiveness of our approach by experimenting on Cambridge Landmarks ( large scale outdoor scenes) as well as 7 Scenes (indoor scenes) using various CNN feature extractors. Our method improves the median error in indoor as well as outdoor localization datasets compared to the previous best deep learning model known as PoseNet (with geometric re-projection loss) using the same feature extractor. We improve the median error in localization in the specific case of Street scene, by over 8m.

Abstract

t Automatic image annotation is one of the fundamental problems in computer vision and machine learning. Given an image, here the goal is to predict a set of textual labels that describe the semantics of that image. During the last decade, a large number of image annotation techniques have been proposed that have been shown to achieve encouragingresults on various annotation datasets. However, their scope has mostly remained restricted to quantitative results on the test data, thus ignoring various key aspects related to dataset properties and evaluation metrics that inherently affect the performance to a considerable extent. In this paper, first we evaluate ten state-of-the-art (both deep-learning based as well as non-deep-learning based) approaches for image annotation using the same baseline CNN features. Then we propose new quantitative measures to examine various issues/aspects in the image annotation domain, such as dataset specific biases, per-label versus per-image evaluation criteria, and the impact of changing the number and type of predicted labels. We believe the conclusions derived in this paper through thorough empirical analyzes would be helpful in making systematic advancements in this domain.

Abstract

In a world of proliferating data, the ability to rapidly summarize text is growing in importance. Automatic summarization of text can be thought of as a sequence to sequence problem. Another area of natural language processing that solves a sequence to sequence problem is machine translation, which is rapidly evolving due to the development of attentionbased encoder-decoder networks. This work applies these modern techniques to abstractive summarization. We perform analysis on various attention mechanisms for summarization with the goal of developing an approach and architecture aimed at improving the state of the art. In particular, we modify and optimize a translation model with self-attention for generating abstractive sentence summaries. The effectiveness of this base model along with attention variants is compared and analyzed in the context of standardized evaluation sets and test metrics. However, we show that these metrics are limited in their ability to effectively score abstractive summaries, and propose a new approach based on the intuition that an abstractive model requires an abstractive evaluation.

Abstract

No Results Found

Abstract

In this paper, we propose a novel method to register football broadcast video frames on the static top view model of the playing surface. The proposed method is fully automatic in contrast to the current state of the art which requires manual initialization of point correspondences between the image and the static model. Automatic registration using existing approaches has been difficult due to the lack of sufficient point correspondences. We investigate an alternate approach exploiting the edge information from the line markings on the field. We formulate the registration problem as a nearest neighbour search over a synthetically generated dictionary of edge map and homography pairs. The synthetic dictionary generation allows us to exhaustively cover a wide variety of camera angles and positions and reduce this problem to a minimal per-frame edge map matching procedure. We show that the per-frame results can be improved in videos using an optimization framework for temporal camera stabilization. We demonstrate the efficacy of our approach by presenting extensive results on a dataset collected from matches of football World Cup 2014.

Abstract

Mobile image retrieval allows users to identify visual information about their environment by transmitting image queries to an online image database that has associated annotations (e.g. location, product information, etc.) with the images. However, this is reliant on a network connection to transmit the query and retrieve the information. This chapter examines mobile image retrieval for offline use when the data net- work connection is limited or not available. In this scenario, the entire visual search index must reside on the mobile device itself. More specifically, we are interested in ‘instance retrieval’, where the annotations associated with the images (e.g. building’s name, object information) are returned by the query and not the images themselves. Figure 1 shows an example use case where mobile camera photos are used to identify buildings and landmarks without the need for a network connection. While our targeted instance retrieval does not need to store the images, the entire visual index structure needs to fit on a mobile device, ideally within a small footprint (e.g. 16–32 MB). This small memory footprint serves two purposes. First, while mobile phones have up to 16–32 GB of storage, this is mainly in the form of slower flash memory that is an order of magnitude slower than RAM. Having the entire index within tens of MBs makes it possible for use in a resident application on the phone’s RAM. Second, this small size is inline with common practices for mobile applications; e.g. the iPhone average application size is currently 23 MB. Addition- ally, iPhone apps less than 50 MB can be downloaded using 3G/4G, anything larger

Abstract

Building bilateral semantic associations between images and texts is among the fundamental problems in computer vision. In this paper, we study two complementary cross-modal prediction tasks: (i) predicting text(s) given a query image (“Im2Text”), and (ii) predicting image(s) given a piece of text (“Text2Im”). We make no assumption on the specific form of text; i.e., it could be either a set of labels, phrases, or even captions. We pose both these tasks in a retrieval framework. For Im2Text, given a query image, our goal is to retrieve a ranked list of semantically relevant texts from an independent text-corpus (i.e., texts with no corresponding images). Similarly, for Text2Im, given a query text, we aim to retrieve a ranked list of semantically relevant images from a collection of unannotated images (i.e., images without any associated textual meta-data). We propose a novel Structural SVM based unified framework for these two tasks, and show how it can be efficiently trained and tested. Using a variety of loss functions, extensive experiments are conducted on three popular datasets (two medium-scale datasets containing few thousands of samples, and one web-scale dataset containing one million samples). Experiments demonstrate that our framework gives promising results compared to competing baseline cross-modal search techniques, thus confirming its efficacy. Keywords: Image search; Image description; Cross-media analysis

Abstract

Human pose as a query modality is an alternative and rich experience for image and video retrieval. It has interesting retrieval applications in domains such as sports and dance databases. In this work we propose two novel ways for representing the image of a person striking a pose, one looking for parts and other looking at the whole image. These representations are then used for retrieval. Both the representations are obtained using deep learning methods. In the first method, we make the following contributions: (a) We introduce ‘deep poselets’ for pose- sensitive detection of various body parts, built on convolutional neural network (CNN) features. These deep poselets significantly outperform previous instantiations of Berkeley poselets [6], and (b) Using these detec- tor responses, we construct a pose representation that is suitable for pose search, and show that pose retrieval performance is on par with the previous methods. In the second method, we make the follow- ing contributions: (a) We design an optimized neural network which maps the input image to a very low dimensional space where similar poses are close by and dissimilar poses are farther away, and (b) We show that pose retrieval system using these low dimensional representation is on par with the deep poselet representation and is on par with the previous methods. The previous works with which the above two methods are compared include bag of visual words [44], Berkeley poselets [6] and human pose estimation algorithms [52]. All the methods are quantitatively eval- uated on a large dataset of images built from a number of standard benchmarks together with frames from

Abstract

Building robust recognizers for Arabic has always been challenging. We demonstrate the effectiveness of an end-to- end trainable CNN - RNN hybrid architecture in recognizing Arabic text in videos and natural scenes. We outperform previous state- of-the-art on two publicly available video text datasets - ALIF and ACTIV . For the scene text recognition task, we introduce a new Arabic scene text dataset and establish baseline results. For scripts like Arabic, a major challenge in developing robust recognizers is the lack of large quantity of annotated data. We overcome this by synthesizing millions of Arabic text images from a large vocabulary of Arabic words and phrases. Our implementation is built on top of the model introduced here [37] which is proven quite effective for English scene text recognition. The model follows a segmentation-free, sequence to sequence transcription approach. The network transcribes a sequence of convolutional features from the input image to a sequence of target labels. This does away with the need for segmenting input image into constituent characters/glyphs, which is often difficult for Arabic script. Further, the ability of RNN s to model contextual dependencies yields superior recognition results.

Abstract

Inspired by the success of Deep Learning based approaches to English scene text recognition, we pose and bench- mark scene text recognition for three Indic scripts - Devanagari, Telugu and Malayalam. Synthetic word images rendered from Unicode fonts are used for training the recognition system. And the performance is bench-marked on a new - IIIT-ILST dataset comprising of hundreds of real scene images containing text in the above mentioned scripts. We use a segmentation free, hybrid but end-to-end trainable CNN - RNN deep neural network for transcribing the word images to the corresponding texts. The cropped word images need not be segmented into the sub- word units and the error is calculated and backpropagated for the the given word image at once. The network is trained using CTC loss, which is proven quite effective for sequence- to-sequence transcription tasks. The CNN layers in the network learn to extract robust feature representations from word images. The sequence of features learnt by the convolutional block is transcribed to a sequence of labels by the RNN + CTC block. The transcription is not bound by word length or a lexicon and is ideal for Indian languages which are highly inflectional.