Faculty Publications

Permanent URI for this communityhttps://idr.nitk.ac.in/handle/123456789/18736

Publications by NITK Faculty

Browse

Author: search.filters.author.Kinnal, B.Has files: No

Search Results

Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    VaFLE: Value flag length encoding for images in a multithreaded environment
    (Springer, 2019) Kinnal, B.; Pasupulety, U.; Geetha, V.
    The Run Length Encoding (RLE) algorithm substitutes long runs of identical symbols with the value of that symbol followed by the binary representation of the frequency of occurrences of that value. This lossless technique is effective for encoding images where many consecutive pixels have similar intensity values. One of the major problems of RLE for encoding runs of bits is that the encoded runs have their lengths represented as a fixed number of bits in order to simplify decoding. The number of bits assigned is equal to the number required to encode the maximum length run, which results in the addition of padding bits on runs whose lengths do not require as many bits for representation as the maximum length run. Due to this, the encoded output sometimes exceeds the size of the original input, especially for input data where in the runs can have a wide range of sizes. In this paper, we propose VaFLE, a general-purpose lossless data compression algorithm, where the number of bits allocated for representing the length of a given run is a function of the length of the run itself. The total size of an encoded run is independent of the maximum run length of the input data. In order to exploit the inherent data parallelism of RLE, VaFLE was also implemented in a multithreaded OpenMP environment. Our algorithm guarantees better compression rates of upto 3X more than standard RLE. The parallelized algorithm attains a speedup as high as 5X in grayscale and 4X in color images compared to the RLE approach. © Springer Nature Singapore Pte Ltd 2019.
  • No Thumbnail Available
    Item
    Performance evaluation of deep learning frameworks on computer vision problems
    (Institute of Electrical and Electronics Engineers Inc., 2019) Nara, M.; Mukesh, B.R.; Padala, P.; Kinnal, B.
    Deep Learning (DL) applications have skyrocketed in recent years and are being applied in various domains. There has been a tremendous surge in the development of DL frameworks to make implementation easier. In this paper, we aim to make a comparative study of GPU-accelerated deep learning software frameworks such as Torch and TenserFlow (with Keras API). We attempt to benchmark the performance of these frameworks by implementing three different neural networks, each designed for a popular Computer Vision problem (MNIST, CIFAR10, Fashion MNIST). We performed this experiment on both CPU and GPU(Nvidia GeForce GTX 960M) settings. The performance metrics used here include evaluation time, training time, and accuracy. This paper aims to act as a guide to selecting the most suitable framework for a particular problem. The special interest of the paper is to evaluate the performance lost due to the utility of an API like Keras and a comparative study of the performance over a user-defined neural network and a standard network. Our interest also lies in their performance when subjected to networks of different sizes. ©2019 IEEE.