Play all

Intro

Image Understanding using context

Finding the Shared Context

Improving understanding on videos usin context • Scene Identification

Image Representation: Top-Down Visual Tree (TDVT)

How to learn TDVT representations?

A. Dataset Parsing (pre-processing)

Inference

A Labels: Top-down Tree LSTM

Model: Top-Down Tree LSTM

B Training 4 Classifiers: Is the node having an X edge? Train 4 classifiers

Evaluation: User study

Object detection Vs proposed approach

Evaluation: Object detection across datasets

Proposed Framework

Words weights for each query image

Incorporation of Context (Random Walk)

Incorporation of Context Random Walk

Experimental setup

Qualitative Results Image Retrieval Application from multiple

Quantitative results

Improving Scene identification on Videos

Improve scene Identification

Object detection on videos using context

Improve Object Detection

Experiments (scene identity)

Experiments (object detection)

Conclusions

Future Work (Image representation)

Description:

Watch a comprehensive thesis defense presentation on image understanding and context utilization in computer vision. Explore innovative approaches to scene identification, object detection, and image representation using Top-Down Visual Tree (TDVT) models. Learn about the incorporation of context through random walk techniques and their application in improving video analysis. Discover experimental results comparing traditional object detection methods with the proposed framework, and gain insights into future research directions in image representation and understanding.

Contextual Image Understanding and Scene Identification in Videos - Thesis Defense

University of Central Florida

Add to list

#Computer Science #Artificial Intelligence #Computer Vision #Machine Learning #Object Detection #Mathematics #Statistics & Probability #Probability Theory #Random Walks