# midas-converter

**Repository Path**: andrew76/midas-converter

## Basic Information

- **Project Name**: midas-converter
- **Description**: Utility to convert RGB frames into depth frames by using MiDas.
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2020-11-15
- **Last Updated**: 2020-12-19

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# MiDas Converter
Utility to convert RGB frames into depth frames by using MiDas.

![](images/office-demo.jpg)

*Office demo image by [fauxels](https://www.pexels.com/@fauxels) from Pexels*

### Installation
It is recommended to use a seperate Python envrionment ([virtualenv](https://virtualenv.pypa.io/en/latest/)) and Python `3.6.X` (Windows 10 & MacOS).

1. Download and install the latest [OpenVINO framework](https://software.intel.com/content/www/us/en/develop/tools/openvino-toolkit.html) (`2021.1`).
2. Run the following script to install the models, convert them and setup the environment:

```bash
# one-script setup of the full environment (grab a coffee ☕)

# windows
setup.bat

# unix
./setup.sh
```

### Demo

Either use a predefined batch script to run the examples or call the demo itself.

```bash
# demo
run.bat
./run.sh

# on your own image
python monodepth_demo.py -m public\midasnet\FP32\midasnet.xml -i yourimage.jpg
```

There is also a live webcam-feed inference example.

```bash
run_webcam.bat
./run_webcam.sh
```

#### Batch Convert

To batch convert a lot of frames run the `monodepth_convert.py` script.

```bash
python monodepth_convert.py -m public\midasnet\FP32\midasnet.xml -i frames
```

### OpenCV Demo

If you are interested in a OpenCV implementation (in Java) feel free to have a look at [the code here](https://github.com/cansik/deep-vision-processing/blob/master/src/main/java/ch/bildspur/vision/MidasNetwork.java#L37-L89).

### About

The [MiDas network](https://github.com/intel-isl/MiDaS) and the pre-trained weights are part of the [Intel ISL](https://github.com/intel-isl). This repository just adds some scripts and utilties to work with it.