Marigold: Repurposing Diffusion-Based Image Generators for Monocular Depth Estimation

This repository represents the official implementation of the paper titled "Repurposing Diffusion-Based Image Generators for Monocular Depth Estimation".

Bingxin Ke, Anton Obukhov, Shengyu Huang, Nando Metzger, Rodrigo Caye Daudt, Konrad Schindler

We present Marigold, a diffusion model and associated fine-tuning protocol for monocular depth estimation. Its core principle is to leverage the rich visual knowledge stored in modern generative image models. Our model, derived from Stable Diffusion and fine-tuned with synthetic data, can zero-shot transfer to unseen data, offering state-of-the-art monocular depth estimation results.

📢 News

2023-12-19: Updated license to Apache License, Version 2.0.
2023-12-08: Added - try it out with your images for free!
2023-12-05: Added - dive deeper into our inference pipeline!
2023-12-04: Added paper and inference code (this repository).

🚀 Usage

We offer several ways to interact with Marigold:

A free online interactive demo is available here: (kudos to the HF team for the GPU grant)
Run the demo locally (requires a GPU and an nvidia-docker2, see Installation Guide): docker run -it -p 7860:7860 --platform=linux/amd64 --gpus all registry.hf.space/toshas-marigold:latest python app.py
Extended demo on a Google Colab:
If you just want to see the examples, visit our gallery:
Finally, local development instructions are given below.

🛠️ Setup

This code was tested on:

Ubuntu 22.04 LTS, Python 3.10.12, CUDA 11.7, GeForce RTX 3090 (pip, Mamba)
CentOS Linux 7, Python 3.10.4, CUDA 11.7, GeForce RTX 4090 (pip)
Windows 11 22H2, Python 3.10.12, CUDA 12.3, GeForce RTX 3080 (Mamba)
MacOS 14.2, Python 3.10.12, M1 16G (pip)

🪟 A Note for Windows users

We recommend running the code in WSL2:

Install WSL following installation guide.
Install CUDA support for WSL following installation guide.
Find your drives in /mnt/<drive letter>/; check WSL FAQ for more details. Navigate to the working directory of choice.

📦 Repository

Clone the repository (requires git):

git clone https://github.com/prs-eth/Marigold.git
cd Marigold

💻 Dependencies

We provide several ways to install the dependencies.

Using Mamba, which can installed together with Miniforge3.

Windows users: Install the Linux version into the WSL.

After the installation, Miniforge needs to be activated first: source /home/$USER/miniforge3/bin/activate.

Create the environment and install dependencies into it:
```
mamba env create -n marigold --file environment.yaml
conda activate marigold
```
Using pip: Alternatively, create a Python native virtual environment and install dependencies into it:
```
python -m venv venv/marigold
source venv/marigold/bin/activate
pip install -r requirements.txt
```

Keep the environment activated before running the inference script. Activate the environment again after restarting the terminal session.

🚀 Testing on your images

📷 Prepare images

If you have images at hand, skip this step. Otherwise, download a few select images from our paper:

bash script/download_sample_data.sh

🎮 Run inference

Place your images in a directory, for example, under input/in-the-wild_example, and run the following command:

python run.py \
    --input_rgb_dir input/in-the-wild_example \
    --output_dir output/in-the-wild_example

You can find all results in output/in-the-wild_example. Enjoy!

⚙️ Inference settings

The default settings are optimized for the best result. However, the behavior of the code can be customized:

Trade-offs between the accuracy and speed (for both options, larger values result in better accuracy at the cost of slower inference.)
- --ensemble_size: Number of inference passes in the ensemble. Default: 10.
- --denoise_steps: Number of denoising steps of each inference pass. Default: 10.
--half_precision: Run with half-precision (16-bit float) to reduce VRAM usage, might lead to suboptimal result.
By default, the inference script resizes input images to the processing resolution, and then resizes the prediction back to the original resolution. This gives the best quality, as Stable Diffusion, from which Marigold is derived, performs best at 768x768 resolution.
- --processing_res: the processing resolution; set 0 to process the input resolution directly. Default: 768.
- --output_processing_res: produce output at the processing resolution instead of upsampling it to the input resolution. Default: False.
--seed: Random seed can be set to ensure additional reproducibility. Default: None (using current time as random seed).
--batch_size: Batch size of repeated inference. Default: 0 (best value determined automatically).
--color_map: Colormap used to colorize the depth prediction. Default: Spectral.
--apple_silicon: Use Apple Silicon MPS acceleration.

⬇ Checkpoint cache

By default, the checkpoint is stored in the Hugging Face cache. The HF_HOME environment variable defines its location and can be overridden:

export HF_HOME=new/path

Alternatively, use the following script to download the checkpoint weights locally:

bash script/download_weights.sh

At inference, specify the checkpoint path:

python run.py \
    --checkpoint checkpoint/Marigold_v1_merged_2 \
    --input_rgb_dir input/in-the-wild_example\
    --output_dir output/in-the-wild_example

✏️ Contributing

Please refer to this instruction.

🤔 Troubleshooting

Problem	Solution
(Windows) Invalid DOS bash script on WSL	Run `dos2unix <script_name>` to convert script format
(Windows) error on WSL: `Could not load library libcudnn_cnn_infer.so.8. Error: libcuda.so: cannot open shared object file: No such file or directory`	Run `export LD_LIBRARY_PATH=/usr/lib/wsl/lib:$LD_LIBRARY_PATH`

🎓 Citation

Please cite our paper:

@misc{ke2023repurposing,
      title={Repurposing Diffusion-Based Image Generators for Monocular Depth Estimation}, 
      author={Bingxin Ke and Anton Obukhov and Shengyu Huang and Nando Metzger and Rodrigo Caye Daudt and Konrad Schindler},
      year={2023},
      eprint={2312.02145},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

🎫 License

This work is licensed under the Apache License, Version 2.0 (as defined in the LICENSE).

By downloading and using the code and model you agree to the terms in the LICENSE.

atumcell / marigold Goto Github PK

marigold's Introduction