This repository contains code to compute depth from a monocular camera (your webcam or Rpi camera) using only OpenCV instead of PyTorch. The code is based on this C++ implementation. The original paper:
Towards Robust Monocular Depth Estimation: Mixing Datasets for Zero-shot Cross-dataset Transfer
René Ranftl, Katrin Lasinger, David Hafner, Konrad Schindler, Vladlen Koltun
MiDaS v2.1 was trained on 10 datasets (ReDWeb, DIML, Movies, MegaDepth, WSVD, TartanAir, HRWSI, ApolloScape, BlendedMVS, IRS) with multi-objective optimization.
I only use MiDaS v2.1 small in this repository, but switching to the full model should be easy to do. You can find the original code here.
Set up dependencies:
conda install opencvYou can use pip if you prefer and you should install the picamera module if you are using a Raspberry PI
pip install "picamera[array]"The code was tested with Python 3.7 and OpenCV 4.5.1.
Run the model:
cd MiDaS_cv2/
python MiDaS.pyMore options:
python3 MiDaS.py -h
usage: MiDaS.py [-h] [--save SAVE] [--rpi_cam] [--no_output] [--normalize]
optional arguments:
-h, --help show this help message and exit
--save SAVE specify a name to save images
--rpi_cam use the raspberry pi camera
--no_output disabling video output
--normalize normalize the output
The following values may vary on your test.
| Speed, FPS | i3-4100M @ 2.50GHz | i5-8350U @ 1.70GHz | RPI 4 @ 1.50GHz | RPI 4 @ 1.75GHz |
|---|---|---|---|---|
| MiDaS v2.1 small | 4.41 | 12.79 | 1.17 | 1.31 |
Please cite this paper if you use this code or any of the models:
@article{Ranftl2020,
author = {Ren\'{e} Ranftl and Katrin Lasinger and David Hafner and Konrad Schindler and Vladlen Koltun},
title = {Towards Robust Monocular Depth Estimation: Mixing Datasets for Zero-shot Cross-dataset Transfer},
journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)},
year = {2020},
}
MIT License