Photo2Depth is a small application that takes as input, an image of a room interior and tries to segment out the furniture in the room and based on the segmentation, generate a similar looking 3D scene, in terms of the relative position of those furniture items. This means its trying to estimate the depth and displacement of the objects, from the viewing camera in the 3D scene, based only on one image input. NOTE: It assumes that all items lie on the same floor level. This project only supports the following furniture items when it comes to detectable furniture in an image: dining tables, chairs, couches, beds, ovens and refrigerators. It allows the user to input their desired image through a basic GUI.
Set up a Python virtual environment by following this guide. Then you can clone this repo and change directory into the main level directory with:
cd individual-masters-projectAfter that, you need to import all the necessary packages. This can be done with:
pip install -r requirements.txtThis will install all the required packages. If for some reason, detectron2 fails to install, try to install it independently with the following command:
python -m pip install 'git+https://github.com/facebookresearch/detectron2.git'For unit testing, the pytest framework was used, so this also needs to be installed in the environment to execute the tests. Once installed with pip, you can run the tests with the following command (assuming your in the individual-masters-project directory):
python -m pytest testsFor more information about usage of pytest, read the official docs.