VBDA_TC_Poster1 - Copy.pdf

Cloud environment node 1 video clips sliding window

Info icon This preview shows page 2. Sign up to view the full content.

Cloud Environment Node-1 Video Clips (Sliding Window) Video Stream Extract Local Feature Descriptors (STIP, HOG, HOF, SIFT) Local Feature Descriptors Node-2 Node-N Master Local Feature Descriptors Distributed Clustering (K-Means) Generate Frequency Histograms Vocabulary (Visual Words) Node-3 Node-i Video Representation (BoVW) Video Representation (BoVW) Data Partitioning (Clustering + Binning) Distributed Learning (BigSVM) Classifier Figure 3: Block diagram of visual big data framework over the cloud for traffic monitoring using large surveillance camera network of a smart city. Depicting use of dis- tributed processing of feature representation using bag-of- words (BoW) and classification using distributed support vector machine (SVM). Implementation Details Ubuntu 16.04 Xenial Xerus 1 Intel(R) Xeon(R) CPU E5-2697 v2 @ 2.70GHz × 48 processor, 128GB RAM, [Tesla K20c] × 2 GPUs 2 Intel(R) Xeon(R) CPU E5-2697 v2 @ 2.70GHz × 16 processor, 64GB RAM, [Tesla K20c] × 6 GPUs C++, OpenCV 3.0, OpenMP and OpenMPI (BoW, k -means and SVM [3]). Result and discussions Table 1: Performance of classification (%) Feature Kernel Bike vs. Non-bike head vs. helmet Linear 98.88 93.80 HOG MLP 82.89 64.50 RBF 82.89 64.50 Linear 82.89 64.51 SIFT MLP 82.89 64.51 RBF 82.89 64.51 Linear 82.89 64.53 LBP MLP 82.89 64.53 RBF 82.89 64.53 - 40 - 20 0 20 40 Principal Component - 1 - 40 - 20 0 20 40 Principal Component - 2 Negative Class Positive Class - 1000 - 500 0 500 1000 Principal Component - 1 - 800 - 600 - 400 - 200 0 200 400 600 800 1000 Principal Component - 2 Negative Class Positive Class (A) (B) Figure 4: Visualization of HOG feature vectors using t- SNE. (A) Bike-rider vs Others. (B) Helmet vs Non- helmet. HOG SIFT LBP 0 20 40 60 80 100 Classification Performance(%) Linear MLP RBF HOG SIFT LBP 0 20 40 60 80 100 Classification Performance(%) Linear MLP RBF (A) (B) Figure 5: Experimental results. (A) Bike-rider vs Others.
Image of page 2
This is the end of the preview. Sign up to access the rest of the document.
  • Fall '16
  • FIX
  • Closed-circuit television, Dinesh Singh, Indian Institute of Technology Hyderabad, visual big data

{[ snackBarMessage ]}

What students are saying

  • Left Quote Icon

    As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

    Student Picture

    Kiran Temple University Fox School of Business ‘17, Course Hero Intern

  • Left Quote Icon

    I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

    Student Picture

    Dana University of Pennsylvania ‘17, Course Hero Intern

  • Left Quote Icon

    The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

    Student Picture

    Jill Tulane University ‘16, Course Hero Intern