Many tasks are related to determining if a particular text string exists in an image. In this work, we propose a new framework that learns this task in an end-to-end way. The framework takes an image and a text string as input and then outputs the probability of the text string being present in the image. This is the first end-to-end framework that learns such relationships between text and images in scene text area. The framework does not require explicit scene text detection or recognition and thus no bounding box annotations are needed. It is also the first work in scene text area that tackles such a weakly labeled problem. Based on this framework, we developed a model called Guided Attention. Our designed model achieves better results than several state-of-the-art scene text reading based solutions for a challenging Street View Business Matching task. The task tries to find correct business names for storefront images and the dataset we collected for it is substantially larger, and more challenging than existing scene text dataset. This new real-world task provides a new perspective for studying scene text related problems.

Many tasks are related to determining if a particular text string exists in an image. In this work, we propose a new framework that learns this task in an end-to-end way. The framework takes an image and a text string as input and then outputs the probability of the text string being present in the image. This is the first end-to-end framework that learns such relationships between text and images in scene text area. The framework does not require explicit scene text detection or recognition and thus no bounding box annotations are needed for it. It is also the first work in scene text area that tackles suh a weakly labeled problem. Based on this framework, we developed a model called Guided Attention. Our designed model achieves much better results than several state-of-the-art scene text reading based solutions for a challenging Street View Business Matching task. The task tries to find correct business names for storefront images and the dataset we collected for it is substantially larger, and more challenging than existing scene text dataset. This new real-world task provides a new perspective for studying scene text related problems. We also demonstrate the uniqueness of our task via a comparison between our problem and a typical Visual Question Answering problem.

We present a neural network model — based on Convolutional Neural Networks, Recurrent Neural Networks and a novel attention mechanism — which achieves 84.2% accuracy on the challenging French Street Name Signs (FSNS) dataset, significantly outperforming the previous state of the art (Smith'16), which achieved 72.46%. Furthermore, our new method is much simpler and more general than the previous approach. To demonstrate the generality of our model, we show that it also performs well on an even more challenging dataset derived from Google Street View, in which the goal is to extract business names from store fronts. Finally, we study the speed/accuracy tradeoff that results from using CNN feature extractors of different depths. Surprisingly, we find that deeper is not always better (in terms of accuracy, as well as speed). Our resulting model is simple, accurate and fast, allowing it to be used at scale on a variety of challenging real-world text extraction problems.

Search with local intent is becoming increasingly useful due to the popularity of the mobile device. The creation and maintenance of accurate listings of local businesses worldwide is time consuming and expensive. In this paper, we propose an approach to automatically discover businesses that are visible on street level imagery. Precise business store front detection enables accurate geo-location of businesses, and further provides input for business categorization, listing generation, etc. The large variety of business categories in different countries makes this a very challenging problem. Moreover, manual annotation is prohibitive due to the scale of this problem. We propose the use of a MultiBox based approach that takes input image pixels and directly outputs store front bounding boxes. This end-to-end learning approach instead preempts the need for hand modeling either the proposal generation phase or the post-processing phase, leveraging large labelled training datasets. We demonstrate our approach outperforms the state of the art detection techniques with a large margin in terms of performance and run-time efficiency. In the evaluation, we show this approach achieves human accuracy in the low-recall settings. We also provide an end-to-end evaluation of business discovery in the real world.

Modern search engines receive large numbers of business related, local aware queries. Such queries are best answered using accurate, up-to-date, business listings, that contain representations of business categories. Creating such listings is a challenging task as businesses often change hands or close down. For businesses with street side locations one can leverage the abundance of street level imagery, such as Google Street View, to automate the process. However, while data is abundant, labeled data is not; the limiting factor is creation of large scale labeled training data. In this work, we utilize an ontology of geographical concepts to automatically propagate business category information and create a large, multi label, training dataset for fine grained storefront classification. Our learner, which is based on the GoogLeNet/Inception Deep Convolutional Network architecture and classifies 208 categories, achieves human level accuracy.

Visual tracking is a challenging problem, as the appearance of an object may change due to viewpoint variations, illumination changes, and occlusion. It may also leave the field of view (FOV), then reappears. In order to track and reacquire an unknown object with limited labeling data, we propose to learn these changes online and incrementally build a model that encodes all appearance variations while tracking. To address this semi-supervised learning problem, we propose a co-training framework with cascade particle filter to label incoming data continuously and online update hybrid generative and discriminative models. Each of the layers in the cascade contains one or more either generative or discriminative appearance models. The cascade manner of organizing the particle filter enables the efficient evaluation of multiple appearance models with different computational costs; thus improves the speed of the tracker. The proposed online framework provides temporally local tracking that adapts to appearance changes. Moreover, it provides an object-specific detection ability that allows to reacquire an object after total occlusion. Extensive experiments demonstrate that under challenging situations, our method has strong reacquisition ability and robustness to distracters in clutter background. We also provide quantitative comparisons to other state of the art trackers.

Detecting pedestrians at a distance from large-format wide-area imagery is a challenging problem because of low ground sampling distance (GSD) and low frame rate of the imagery. In such a scenario, the approaches based on appearance cues alone mostly fail because pedestrians are only a few pixels in size. Frame-differencing and optical flow based approaches also give poor detection results due to noise, camera jitter and parallax in aerial videos. To overcome these challenges, we propose a novel approach to extract Multi-scale Intrinsic Motion Structure features from pedestrian's motion patterns for pedestrian detection. The MIMS feature encodes the intrinsic motion properties of an object, which are location, velocity and trajectory-shape invariant. The extracted MIMS representation is robust to noisy flow estimates. In this paper, we give a comparative evaluation of the proposed method and demonstrate that MIMS outperforms the state of the art approaches in identifying pedestrians from low resolution airborne videos.

In this work, we propose a novel video representation for activity recognition that models video dynamics with attributes of activities. A video sequence is decomposed into short-term segments, which are characterized by the dynamics of their attributes. These segments are modeled by a dictionary of attribute dynamics templates, which are implemented by a recently introduced generative model, the binary dynamic system~(BDS). We propose methods for learning a dictionary of BDSs from a training corpus, and for quantizing attribute sequences extracted from videos into these BDS code words. This procedure produces a representation of the video as a histogram of BDS code words, which is denoted the bag-of-words for attribute dynamics (BoWAD). An extensive experimental evaluation reveals that this representation outperforms other state-of-the-art approaches in temporal structure modeling for complex activity recognition.

In this paper, we present a interactive Event Detection (interactiveED) system evaluated in TRECVID 2013 competition. The proposed system is evaluated on all the seven event categories. Two channels of raw features: STIP-HOG/HOF and SURF/MHI-HOG ‐ are extracted from down-sampled surveillance video frames. A Bag Of Words model is used to generate representations of segments of surveillance videos. Segments of videos are sampled using sliding windows. Spatial priors are utilized in two channels: Hot Region and Human Tracking Masks. Spatial priors are used for filtering noisy feature responses. Feature mapping technique is employed to generate more features, which potentially increase the separability of sample set in feature space. A cascade of SVMs are learned to tackle the classification task, with respect to different camera views. Human interaction is utilized as a post-processing layer to filter out false alarms. A Graphical User Interface (GUI) is developed to help an expert user to accomplish the task efficiently and conveniently.

In Multimedia Event Detection 2013 evaluation, SRI Aurora team participated in EK100, EK10, and EK0 tasks with full system evaluation. We submitted 15 runs for both pre-specified events (PS-Events) and ad-hoc events (AH-Events). The majority of them achieved satisfactory results. In particular, thanks to the well-designed concept features, our EK10 system works consistently much better for both PSEvents and AH-Events. By creating the concept language model from the web source, we build our EK0 system to perform event detection without training examples. This system achieved promising results on PS-Events. In MER task, we developed an approach to provide a breakdown of the evidences of why the MED decision has been made by exploring the SVM-based event detector. Furthermore, we designed evidence specific verification and detection to reduce uncertainty and improve key evidence discovery.

We propose to use action, scene and object concepts as semantic attributes for classification of video events in InTheWild content, such as YouTube videos. We model events using a variety of complementary semantic attribute features developed in a semantic concept space. Our contribution is to systematically demonstrate the advantages of this concept-based event representation (CBER) in applications of video event classification and understanding. Specifically, CBER has better generalization capability, which enables to recognize events with a few training examples. In addition, CBER makes it possible to recognize a novel event without training examples (i.e., zero-shot learning). We further show our proposed enhanced event model can further improve the zero-shot learning. Furthermore, CBER provides a straightforward way for event recounting/understanding. We use the TRECVID Multimedia Event Detection (MED11) open source event definitions and datasets as our test bed and show results on over 1400 hours of videos.

The problem of adaptively selecting pooling regions for the classification of complex video events is considered. Complex events are defined as events composed of several characteristic behaviors, whose temporal configuration can change from sequence to sequence. A dynamic pooling operator is defined so as to enable a unified solution to the problems of event specific video segmentation, temporal structure modeling, and event detection. Video is decomposed into segments, and the segments most informative for detecting a given event are identified, so as to dynamically determine the pooling operator most suited for each sequence. This dynamic pooling is implemented by treating the locations of characteristic segments as hidden information, which is inferred, on a sequence-by-sequence basis, via a large-margin classification rule with latent variables. Although the feasible set of segment selections is combinatorial, it is shown that a globally optimal solution to the inference problem can be obtained efficiently, through the solution of a series of linear programs. Besides the coarse-level location of segments, a finer model of video structure is implemented by jointly pooling features of segment-tuples. Experimental evaluation demonstrates that the resulting event detector has state-of-the-art performance on challenging video datasets.

In this paper, we describe the evaluation results for TRECVID 2012 Multimedia Event Detection (MED) and Multimedia Event Recounting (MER) tasks as a part of SRI-Sarnoff AURORA system that is developed under the IARPA ALDDIN program. In AURORA system, we incorporated various low-level features that capture color, appearance, motion, and audio information in videos. Based on these low-level features, we developed Fixed-Pattern and Object-Orientated spatial feature pooling, which result in significant performance improvement to our system. In addition, we collected more than 1800 concepts and designed a set of concept pooling approaches to build the Concept Based Event Representation (CBER, i.e., high-level features). We submitted six runs exploring various fusions of low-level features, high-level features, and ASR/OCR features for MED task. All runs achieve satisfactory results. In particular, two EK10Ex runs for both pre-specified events (PS-Events) and ad-hoc events (AH-Events) obtain relatively better results. In MER task, we developed an approach to provide a breakdown of the evidences of why the MED decision has been made by exploring the SVM-based event detector. Furthermore, we designed evidence specific verification and detection to reduce uncertainty and improve key evidence discovery. Our MER evaluation results for MER-to-Event are very good.

Multimedia event detection has drawn a lot of attention in recent years. Given a recognized event, in this paper, we conduct a pilot study of the multimedia event recounting problem, which answers the question why this video is recognized as this event, i.e. what evidences this decision is made on. In order to provide a semantic recounting of the multimedia event, we adopt a concept-based event representation for learning a discriminative event model. Then, we present a recounting approach that exactly recovers the contribution of semantic evidence to the event classification decision. This approach can be applied on any additive discriminative classifiers. The promising result is shown on the MED11 dataset that contains 15 events in thousands of YouTube like videos.