Abstract

Visual place recognition on low memory devices such as mobile phones and robotics systems is a challenging problem. The state of the art models for this task uses deep learning architectures having close to 100 million parameters which takes over 400MB of memory. This makes these models infeasible to be deployed in low memory devices and gives rise to the need of compressing them. Hence we study the effectiveness of model compression techniques like trained quantization and pruning for reducing the number of parameters on one of the best performing image retrieval models called NetVLAD. We show that a compressed network can be created by starting with a model pre-trained for the task of visual place recognition and then fine-tuning it via trained pruning and quantization. The compressed model is able to produce the same mAP as the original uncompressed network. We achieve almost 50% parameter pruning with no loss in mAP and 70% pruning with close to 2% mAP reduction, while also performing 8-bit quantization. Furthermore, together with 5-bit quantization, we perform about 50% parameter reduction by pruning and get only about 3% reduction in mAP. The resulting compressed networks have sizes of around 30MB and 65MB which makes them easily usable in memory constrained devices.

Abstract

Building robust text recognition systems for languages with cursive scripts like Urdu has always been challenging. Intricacies of the script and the absence of ample annotated data further act as adversaries to this task. We demonstrate the effectiveness of an end-to-end trainable hybrid CNN-RNN architecture in recognizing Urdu text from printed documents, typically known as Urdu OCR. The solution proposed is not bounded by any language specific lexicon with the model following a segmentation-free, sequence-tosequence transcription approach. The network transcribes a sequence of convolutional features from an input image to a sequence of target labels. This discards the need to segment the input image into its constituent characters/glyphs, which is often arduous for scripts like Urdu. Furthermore, past and future contexts modelled by bidirectional recurrent layers aids the transcription. We outperform …

Abstract

This paper proposes a method for detecting plagiarism in source-codes using deep features. The embeddings for programs are obtained using a character-level Recurrent Neural Network (char-RNN), which is pre-trained on Linux Kernel source-code. Many popular plagiarism detection tools are based on n-gram techniques at syntactic level. However, these approaches to plagiarism detection fail to capture long term dependencies (non-contiguous interaction) present in the source-code. Contrarily, the proposed deep features capture non-contiguous interaction within n-grams. These are generic in nature and there is no need to fine-tune the char-RNN model again to program submissions from each individual problem-set. Our experiments show the effectiveness of deep features in the task of classifying assignment program submissions as copy, partial-copy and non-copy. Comparing our proposed features with …

Abstract

While the problem of detecting generic objects in natural scene images has been the subject of research for a long time, the problem of detection of small objects has been largely ignored. While generic object detectors perform well on medium and large sized objects, they perform poorly for the overall task of recognition of small objects. This is because of the low resolution and simple shape of most small objects. In this work, we suggest a simple yet effective upsampling-based technique that performs better than the current state-of-the-art for end-to-end small object detection. Like most recent methods, we generate proposals and then classify them. We suggest improvements to both these steps for the case of small objects.

Abstract

e datasets. We propose an EEG (Electroencephalogram)-based image annotation system. While humans can recognize objects in 20–200 ms, the need to manually label images results in a low annotation throughput. Our system employs brain signals captured via a consumer EEG device to achieve an annotation rate of up to 10 images per second. We exploit the P300 event-related potential (ERP) signature to identify target images during a rapid serial visual presentation (RSVP) task. We further perform unsupervised outlier removal to achieve an F1-score of 0.88 on the test set. The proposed system does not depend on category-specific EEG signatures enabling the annotation of any new image category without any model pre-training.

Abstract

In this paper, we demonstrate a score based indexing approach for tennis videos. Given a broadcast tennis video (btv), we index all the video segments with their scores to create a navigable and searchable match. Our approach temporally segments the rallies in the video and then recognizes the scores from each of the segments, before refining the scores using the knowledge of the tennis scoring system. We finally build an interface to effortlessly retrieve and view the relevant video segments by also automatically tagging the segmented rallies with human accessible tags such as ‘fault’ and ‘deuce’. The efficiency of our approach is demonstrated on btv’s from two major tennis tournaments.

Abstract

In a network of n nodes (modelled as a digraph), the goal of a perfectly secret message transmission (PSMT) protocol is to replicate sender’s message m at the receiver’s end without revealing any information about m to a computationally unbounded adversary that eavesdrops on any t nodes. The adversary may be mobile too – that is, it may eavesdrop on a different set of t nodes in different rounds. We prove a necessary and sufficient condition on the synchronous network for the existence of r-round PSMT protocols, for any given r > 0; further, we show that round-optimality is achieved without trading-off the communication complexity; specifically, our protocols have an overall communication complexity of O(n) elements of a finite field to perfectly transmit one field element. Apart from optimality/scalability, two interesting implications of our results are: (a) adversarial mobility does not affect its tolerability: PSMT tolerating a static tadversary is possible if and only if PSMT tolerating mobile t-adversary is possible; and (b) mobility does not affect the round optimality: the fastest PSMT protocol tolerating a static t-adversary is not faster than the one tolerating a mobile t-adversary.

Abstract

The ideal goal of any program logic is to develop logically correct programs without the need for predominant debugging. Separation logic is considered to be an effective program logic for proving programs that involve pointers. Reasoning with pointers becomes difficult especially due to the way they interfere with the modular style of program development. For instance, often there is aliasing arising due to several pointers to a given cell location. In such situations, any alteration to that cell in one of the program modules may affect the values of many syntactically unrelated expressions in other modules. In this chapter, we try to explore these difficulties through some simple examples and introduce the notion of separating conjunction as a tool to deal with it. We introduce separation logic as an extension of Hoare Logic using a programming language that has four pointer-manipulating commands. These commands perform the usual heap operations such as lookup, update, allocation and de-allocation. The new set of assertions and axioms of separation logic are presented in a semi-formal style. Examples are given to illustrate unique features of the new assertions and axioms. Finally, the chapter concludes with proofs of some real programs using the axioms of separation logic.

Abstract

No Results Found

Abstract

In this paper, utilizing the feature of the sonority, the problems of detecting the syllable nuclei and the syllable boundaries are explored. As there is minimal obstruction in the oral cavity in the production of sonorant sounds, sonority feature is used. A method is proposed for extracting the sonority profile from the strength of the dominant resonance frequency using the zero time windowing method. The strength of the dominant resonance frequencies in the specific band of 500 Hz to 1700 Hz is used. The proposed method is evaluated on SVL-DD-NEWScorpora. The results are compared with the Fourier transform method and the energy-based method of extracting the sonority profile. The proposed method performed better compared to other methods, especially in the presence of noise such as white, pink and babble. The method is also evaluated on TIMIT test database and the performance is on par with the current methods in detecting the syllable nuclei.