1 / 18

Project 2: Automatic Image Labeling

Project 2: Automatic Image Labeling. Presented by: Sadia Anwar Sara Tily Seth Wessitsh. CSC 864 Final Project Presentation. Mid Term Presentation. Overview. Training Data Database of annotated images Words associated with every segment in training data

oona
Télécharger la présentation

Project 2: Automatic Image Labeling

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. Project 2: Automatic Image Labeling Presented by: Sadia Anwar Sara Tily Seth Wessitsh CSC 864 Final Project Presentation

  2. Mid Term Presentation

  3. Overview • Training Data • Database of annotated images • Words associated with every segment in training data • Automatic Image Annotation • Automatic association of words in the training data to the segments of the image CSC 864 Mid Term Presentation

  4. Training Image Data Original Image JSEG JSEG Segmented Image Manually Annotate Program Calculates Image ID Segment ID Labels Feature Vectors CSC 864 Mid Term Presentation

  5. Image Annotation Original Image JSEG Segmented Image Program Performs <- Comparison -> Training Data Vectors Feature Vectors CSC 864 Mid Term Presentation

  6. Result

  7. Final Presentation

  8. Implemented Features • Color (HSI) • Shape Recognition • Rectangularity R = Ao /AR Ao Area of a segment AR Area of smallest rectangular region around that segment • Relative Area • Relative Position

  9. Training Data Training Image JSEG JSEG Segmented Image Manually Annotate Program Calculates Image ID Segment ID Labels Feature Vectors Program Calculates Mean Segments CV and SD Vector

  10. Formulation of Training Data • Mean segments are computed by taking mean values of the features for segments with the same labels • Standard Deviation is computed similarly • Coefficient of variation is calculated for each feature of the mean segments CV = σ/Mean

  11. Un-annotated Image JSEG Segmented Image Program Performs <- Comparison -> and assign labels Training Data Vectors Feature Vectors Image Annotation

  12. Associating Segments With Words [1] • For every segment, compare its feature vector with all the representative feature vectors in the training image set • Comparison is made by computing similarity vector which is the cosine distance between each feature • Multiply CV vector and the similarity vector • Compute magnitude of the similarity vector to get a single value

  13. Associating Segments With Words [2] • Threshold is computed as a distance between the representative vector and the SD vector • If (similarity) < threshold && (similarity) < similarityprevious assign the word to the segment

  14. Results [1] Neither CV nor T CV only

  15. Results [2] T Only CV and T

  16. Results [3] Neither CV nor T CV only

  17. Results [4] T Only Both CV and T

  18. Experimental Evaluation

More Related