Project Description
Overview and Aim
The aim of this project is to develop a system that leverages historical photographs from the fotoatlas.net archive. These photographs are typically annotated manually to highlight important objects or landmarks. In this project, students will create an AI-driven system to automate that annotation process and build a search API. This will allow users to search the archive in two ways: by uploading a modern photo to find historical matches of the same location or object, and by searching for a specific object using textual descriptions to see where it appears in the archive. In other words, it’s a visual object search that combines AI-based annotation and an intuitive photo-driven search capability.
Additionally, the project will include an interface where a human reviewer can oversee the automatically generated annotations. This means a person will be able to approve, reject, or correct the AI-generated tags, ensuring a layer of human validation and adjustment when needed. This human-in-the-loop feature will help maintain accuracy and improve the overall quality of the annotations.
Deliverables:
- An AI-powered system that automatically annotates historical photographs from the fotoatlas.net archive.
- A human-in-the-loop interface allowing reviewers to approve, reject, or correct the automated annotations.
- A visual object search capability that lets users upload a modern photo and find historical matches or search for specific objects across the archive.
- A fully functional API server enabling these search features, making the annotated archive accessible and searchable through both image-based and keyword-based queries.
Sample foto from fotoatlas for Arabali Vapur:
Please explore the site to gain a better understanding.
Collaboration with Dr. Onur Güngör
Project Description
Overview and Aim
The aim of this project is to develop a system that leverages historical photographs from the fotoatlas.net archive. These photographs are typically annotated manually to highlight important objects or landmarks. In this project, students will create an AI-driven system to automate that annotation process and build a search API. This will allow users to search the archive in two ways: by uploading a modern photo to find historical matches of the same location or object, and by searching for a specific object using textual descriptions to see where it appears in the archive. In other words, it’s a visual object search that combines AI-based annotation and an intuitive photo-driven search capability.
Additionally, the project will include an interface where a human reviewer can oversee the automatically generated annotations. This means a person will be able to approve, reject, or correct the AI-generated tags, ensuring a layer of human validation and adjustment when needed. This human-in-the-loop feature will help maintain accuracy and improve the overall quality of the annotations.
Deliverables:
Sample foto from fotoatlas for
Arabali Vapur:Please explore the site to gain a better understanding.
Collaboration with Dr. Onur Güngör