A novel photography support system for PVI called VisPhoto. The first key idea is to use an omnidirectional camera to take a photograph. It allows the user not to take time to aim the camera at the target. However, the captured image includes many objects and a large amount of background that the user does not want to include in the photograph. Therefore, as the second key idea, in post-production, VisPhoto generates a photograph by outputting the cropped region that the user wants to include in the photograph. To help the user to determine which region to crop, our system applies an object detection technique to the image and provides the user with information on which objects exist in which direction.
The structure of this project is shown below.
The src
directory contains 2 apps: sync
and web
.
.
├── README.md
├── docker
├── docker-compose.yml
└── src
├── sync
└── web
sync
: batch application that periodically downloads omnidirectional camera images from Google Drive along with object detection and speech recognition results.
web
: web application to crop the user's desired area from the omnidirectional camera image.
The applications depend on Google Cloud API. Thus, you should prepare your own API credentaial before starting.
In addition, sync
app uses Rclone to download files from Google Drive, so you need to set up rclone.conf
for your Google account.
- Clone this repo to your local machine.
% git clone [email protected]:opu-imp/VisPhoto-ASSETS-2023-tfcam.git visphoto-webapp
- Place your Google Cloud API credentails.
$ cp your-api-secret.json /src/sync/auth/VisPhoto-537053e46c2f.json
$ cp your-api-secret.json /src/web/auth/VisPhoto-537053e46c2f.json
- Place your
rclone.conf
.
$ mv rclone.conf src/sync/rclone
- Launch the application.
% cd visphoto-webapp
% docker compose up -d
- You can access the site from http://localhost:8080