1 / 9

Action Recognition

Action Recognition. Karthik Prabhakar UCF REU 2008, Week 1 Report May 23, 2008. About Me. Purdue University – Indianapolis Computer Science ( Junior Senior) Non-Academic Interests: Cricket. Major Research Projects.

cuyler
Download Presentation

Action Recognition

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Action Recognition Karthik Prabhakar UCF REU 2008, Week 1 Report May 23, 2008

  2. About Me • Purdue University – Indianapolis • Computer Science (Junior Senior) • Non-Academic Interests: • Cricket

  3. Major Research Projects • “Identifying and Extracting Biological Entity Relationships From Textual Documents Using Support Vector Machines” • March 2006 – July 2007 • Developed a system to extract biological relationships (gene-gene, protein-protein, etc.) in free-flowing biomedical text abstracts. • “Real-Time Mental Workload Evaluation Using Multi-Modal Biometric Data” • August 2007 – January 2008 • Designed a real-time eye tracking software for the integrated eye tracking device. • Designed and implemented a system to evaluate mental workload based on eye characteristics (movement, pupil dilation) and facial thermal images

  4. Individual ‘Research’ Projects • Face Recognition Using PCA and LDA • Yale Face Database B (15 subjects, 11 images per) • PCA  SVM + LDA  SVM • Optimal: 2 training per = 68% accuracy on rank 1 • Multimodal Biometric System Using Face and Iris Recognition • Combined face recognition from above with iris recognition • Combined classifier outputs from both. • Optimal: 2 training face, 1 training iris = 78% accuracy on rank 1 • Object Recognition Using Local and Global Features • Again, combination at classifier outputs. • Caltech101 database. • 12 classes, 50 images per (20 training, 30 testing). Accuracy = 31%

  5. Future Plans • Goals and expectations for the summer: • Experience research at another institution. • Contribute positively to the program. • Publish a paper. • Long-term plans: • Graduate School (PhD in Computer Science) • Become a professor….hopefully!

  6. Edge Detection (Video) Canny Filter s = 1.6, L = 0.5, H = 0.8 Original Video Default Canny Filter

  7. Edge Detection (People Count)

  8. Segmentation • Segmentation results using k-means: Original K-means (color) K-means (color + dist)

  9. Segmentation (contd.) Original K-means (color) K-means (color + dist)

More Related