(Explainable) Data-Centric AI: what are you explaininhg, and to whom?
Inside prototype 16-113
1. VIDEO SEARCHING BY AUTOMATIC ANNOTATION
GROUP NUMBER : 16 - 113
INSIDEPROTOTYPE
2. “INSIDE” a smart semantic video
searching server side web application with
automatic annotation which support easy,
quick user friendly querying for an
accurate result
3. Video structure analysis
Analyze the video in order to identify the shot
boundaries separately by fragmenting the video
into frames and identify duplicate frames
4. Video structure analysis (done so far)
• Fragment the video into frames
• Identify duplicate images using hashing algorithm
5. Video structure analysis (to be done)
• Optimize the code to increase the accuracy
• Use better techniques with hashing algorithm to identify duplicates
7. Deep Learning (done so far)
• Creation of the model using Tensorflow machine learning framework
• Classify the objects into categories using the model
• Storing the identified objects in a specific video in to a file
10. Data set preparation and Background Identification (done so far)
• Categorizing the images into classes
• Creation of the file that map the classes string and the UID(unique
id used by the model)
• Creation of the file that map the UID and the labels of the images
11. Data set preparation and Background Identification (done so far)
• Identify the background of the fragmented frames of the video
clip
13. Semantic textual searching (done so far)
• Ontology creation
• Tokenizing the user query and match the tokens with the ontology
and get a relationship
• Match the relationship with the database objects and rank the
videos
14. Semantic textual searching (to be done)
• Analyzing the relationships and categorize videos automatically