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.

Abstract

This paper explores the effectiveness of statistical language model ( SLM ) and dictionary based methods for detec- tion and correction of errors in Indic OCR output. In SLM , we use unicode level ngrams for building the language model. We compare its performance with akshara level ngrams and find that akshara level ngrams perform better in detecting the errors when compared to unicode level ngrams. We experimentally analyze the performance of Indic OCR post-processing using dictionary method, compare the performance with English and analyze the reasons for the under-performance in Indic scripts. We use four major Indian languages for our experiments, namely Hindi, Gurumukhi, Telugu and Malayalam.

Abstract

The power of ConvNets has been demonstrated in a wide variety of vision tasks including pose estimation. But they often produce absurdly erroneous predictions in videos due to unusual poses, challenging illumination, blur, self-occlusions etc. These erroneous predictions can be re- fined by leveraging previous and future predictions as the temporal smoothness constrain in the videos. In this paper, we present a generic approach for pose correction in videos using sequence learning that makes minimal assumptions on the sequence structure. The proposed model is generic, fast and surpasses the state-of-the-art on benchmark datasets. We use a generic pose estimator for initial pose estimates, which are further refined using our method. The proposed architecture uses Long Short-Term Memory ( LSTM ) encoder- decoder model to encode the temporal context and refine the estimations. We show 3.7% gain over the baseline Yang & Ramanan ( YR ) [1] and 2.07% gain over Spatial Fusion Network ( SFN ) [2] on a new challenging YouTube Pose Subset dataset [3]. Keywords-Pose estimation; sequence to sequence learnin

Abstract

We look at the problem of developing a compact and accurate model for gesture recognition from videos in a deep-learning framework. Towards this we propose a joint 3DCNN-LSTM model that is end-to-end trainable and is shown to be better suited to capture the dynamic information in actions. The solution achieves close to state-of-the-art accuracy on the ChaLearn dataset, with only half the model size. We also explore ways to derive a much more compact representation in a knowledge distillation framework followed by model compression. The final model is less than 1 MB in size, which is less than one hundredth of our initial model, with a drop of 7% in accuracy, and is suitable for real-time gesture recognition on mobile devices.

Abstract

We will present our work on controlling nuclear and electronic motion in molecules using laser pulses designed using Optimal control theory (OCT). The control problem is formulated as the maximization of cost functional made of an objective function along with the constraints on the field fluence. The objective function has the prescribed dynamical goals. A detailed analysis of the optimal pulse reveals complex path-ways to arrive at seemingly simple goals. We have used OCT to obtain infrared laser pulses for the selective vibrational excitation of several different systems, from a simple diatomic to triatomic systems located in complex molecular environments. Controlled initiation of intramolecular H-transfer in malonaldehyde using ultrashort uv-laser pulses through an optically bright electronic excited state as a mediator will be discussed. Finally, our attempts to control the quantum dynamics of N-H photodissociation of pyrrole via a counterintuitive coupling of electronic states using uv-laser pulses will be discussed.

Abstract

A first principles quantum dynamics study of N–H photodissociation of pyrrole on the coupled electronic states is carried out with the aid of an optimally designed UV-laser pulse. A new photodissociation path, as compared to the conventional barrier crossing on the state, opens up upon electronic transitions under the influence of pump-dump laser pulses, which efficiently populate both the dissociation channels. The interplay of electronic transitions due both to vibronic coupling and the laser pulse is observed in the control mechanism and discussed in detail. The proposed control mechanism seems to be robust, and not discussed in the literature so far, and is expected to trigger future experiments on the photochemistry of molecules of chemical and biological importance. The design of the optimal pulses and their application to enhance the overall dissociation probability is carried out within the framework of optimal control theory. The quantum dynamics of the system in the presence of pulse is treated by solving the time-dependent Schr ̈ odinger equation in the semi-classical dipole approximation

Abstract

Optimal initiation of quantum dynamics of N-H photodissociation of pyrrole on the S0- 1πσ∗( 1A2) coupled electronic states by UV-laser pulses in an effort to guide the subsequent dynamics to dissociation limits is studied theoretically. Specifically, the task of designing optimal laser pulses that act on initial vibrational states of the system for an effective UV-photodissociation is considered by employing optimal control theory. The associated control mechanism(s) for the initial state dependent photodissociation dynamics of pyrrole in the presence of control pulses is examined and discussed in detail. The initial conditions determine implicitly the variation in the dissociation probabilities for the two channels, upon interaction with the field. The optimal pulse corresponds to the objective fixed as maximization of overall reactive flux subject to constraints of reasonable fluence and quantum dynamics. The simple optimal pulses obtained by the use of genetic algorithm based optimization are worth an experimental implementation given the experimental relevance of πσ∗-photochemistry in recent times.

Abstract

There are numerous situations when a service requester wishes to expertsourcea series of identical but non-trivial tasks from a pool of experts so as to achieve an assured accuracy level for each task, in a cost optimal way. The experts available are typically heterogeneous with unknown but fixed qualities and different service costs. The service costs are usually private to the experts and the experts could be strategic about their costs. The problem is to select for each task an optimal subset of experts so that the outcome obtained after aggregating the opinions from the selected experts guarantees a target level of accuracy. The problem is a challenging one even in a non-strategic setting since the accuracy of an aggregated outcome depends on unknown qualities. We develop a novel multi-armed bandit (MAB) mechanism for solving this problem. First, we propose a framework, Assured Accuracy Bandit (AAB)framework, which leads to a MAB algorithm, Constrained Confidence Bound for Non-Strategic Setting (CCB-NS). We derive an upper bound on the number of time steps this algorithm chooses a sub-optimal set, which depends on the target accuracy and true qualities. A more challenging situation arises when the requester not only has to learn the qualities of the experts but has to elicit their true service costs as well. We modify the CCB-NS algorithm to obtain an adaptive exploration separated algorithm Constrained Confidence Bound for Strategic Setting (CCB-S). The CCB-S algorithm produces an ex-post monotone allocation rule that can then be transformed into an ex-post incentive compatible and ex-post individually rational mechanism. This mechanism learns the qualities of the experts and guarantees a given target accuracy level in a cost optimal way. We also provide a lower bound on the number of times any algorithm must select a sub-optimal set and we see that the lower bound matches our upper bound up to a constant factor. We provide insights on a practical implementation of this framework through an illustrative example and demonstrate the efficacy of our algorithms through simulations