Skip to content

Latest commit

 

History

History
57 lines (39 loc) · 1.46 KB

ppt.md

File metadata and controls

57 lines (39 loc) · 1.46 KB
marp theme paginate backgroundColor color header footer
false
gaia
true
Clash of T-AI-TANS
Lil James @ Saras AI Institute

Clash of T-AI-TANS

Advanced Image Retrieval System
Presented by:
Lil James @ Saras AI Institute

Introduction

Objective:

  • Develop an advanced image retrieval system that enables users to search for images based on textual descriptions.

System Overview:

  • Uses pre-trained models to understand and match the content of images with user-provided descriptions.

Real-Life Applications (1/2)

1. Education:

  • Problem: Difficulty in indexing and searching images in class notes.
  • Solution: Text-based search for students to easily find images related to specific topics.

Real-Life Applications (2/2)

2. Satellite Imaging:

  • Problem: Challenges in identifying key features in vast satellite images.
  • Solution: Search for events like landslides or tsunami survivors using descriptive text.

High-Level Implementation (1/2)

1. Image Captioning Model:

  • Uses pre-trained models (e.g., Git-Large by Microsoft) to generate text descriptions for images.

2. Text Embedding Model:

  • Utilizes a model like BERT to convert descriptions into vectors, allowing similarity search through cosine similarity.

High-Level Implementation (2/2)

3. Vector Database:

  • Stores embedding vectors to reduce computational load and hardware requirements for queries.