{"id":13958454,"url":"https://github.com/isl-org/MiDaS","last_synced_at":"2025-07-21T00:30:55.058Z","repository":{"id":38411299,"uuid":"193518067","full_name":"isl-org/MiDaS","owner":"isl-org","description":"Code for robust monocular depth estimation described in \"Ranftl et. al., Towards Robust Monocular Depth Estimation: Mixing Datasets for Zero-shot Cross-dataset Transfer, TPAMI 2022\"","archived":false,"fork":false,"pushed_at":"2024-08-23T10:21:06.000Z","size":3471,"stargazers_count":4910,"open_issues_count":162,"forks_count":672,"subscribers_count":70,"default_branch":"master","last_synced_at":"2025-05-21T13:11:18.684Z","etag":null,"topics":["deeplearning","monocular-depth-estimation","single-image-depth-prediction"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/isl-org.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":"SECURITY.md","support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2019-06-24T14:08:35.000Z","updated_at":"2025-05-21T04:14:00.000Z","dependencies_parsed_at":"2024-01-16T06:03:30.474Z","dependency_job_id":"2c264e1e-ecbb-4498-9203-f1552396b143","html_url":"https://github.com/isl-org/MiDaS","commit_stats":{"total_commits":74,"total_committers":15,"mean_commits":4.933333333333334,"dds":0.4054054054054054,"last_synced_commit":"bdc4ed64c095e026dc0a2f17cabb14d58263decb"},"previous_names":[],"tags_count":5,"template":false,"template_full_name":null,"purl":"pkg:github/isl-org/MiDaS","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/isl-org%2FMiDaS","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/isl-org%2FMiDaS/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/isl-org%2FMiDaS/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/isl-org%2FMiDaS/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/isl-org","download_url":"https://codeload.github.com/isl-org/MiDaS/tar.gz/refs/heads/master","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/isl-org%2FMiDaS/sbom","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":266221247,"owners_count":23894964,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["deeplearning","monocular-depth-estimation","single-image-depth-prediction"],"created_at":"2024-08-08T13:01:36.358Z","updated_at":"2025-07-21T00:30:54.079Z","avatar_url":"https://github.com/isl-org.png","language":"Python","funding_links":[],"categories":["其他_机器视觉","Depth Estimation"],"sub_categories":["网络服务_其他","Classic Models"],"readme":"## Towards Robust Monocular Depth Estimation: Mixing Datasets for Zero-shot Cross-dataset Transfer\n\nThis repository contains code to compute depth from a single image. It accompanies our [paper](https://arxiv.org/abs/1907.01341v3):\n\n\u003eTowards Robust Monocular Depth Estimation: Mixing Datasets for Zero-shot Cross-dataset Transfer  \nRené Ranftl, Katrin Lasinger, David Hafner, Konrad Schindler, Vladlen Koltun\n\n\nand our [preprint](https://arxiv.org/abs/2103.13413):\n\n\u003e Vision Transformers for Dense Prediction  \n\u003e René Ranftl, Alexey Bochkovskiy, Vladlen Koltun\n\nFor the latest release MiDaS 3.1, a [technical report](https://arxiv.org/pdf/2307.14460.pdf) and [video](https://www.youtube.com/watch?v=UjaeNNFf9sE\u0026t=3s) are available.\n\nMiDaS was trained on up to 12 datasets (ReDWeb, DIML, Movies, MegaDepth, WSVD, TartanAir, HRWSI, ApolloScape, BlendedMVS, IRS, KITTI, NYU Depth V2) with\nmulti-objective optimization. \nThe original model that was trained on 5 datasets  (`MIX 5` in the paper) can be found [here](https://github.com/isl-org/MiDaS/releases/tag/v2).\nThe figure below shows an overview of the different MiDaS models; the bubble size scales with number of parameters.\n\n![](figures/Improvement_vs_FPS.png)\n\n### Setup \n\n1) Pick one or more models and download the corresponding weights to the `weights` folder:\n\nMiDaS 3.1\n- For highest quality: [dpt_beit_large_512](https://github.com/isl-org/MiDaS/releases/download/v3_1/dpt_beit_large_512.pt)\n- For moderately less quality, but better speed-performance trade-off: [dpt_swin2_large_384](https://github.com/isl-org/MiDaS/releases/download/v3_1/dpt_swin2_large_384.pt)\n- For embedded devices: [dpt_swin2_tiny_256](https://github.com/isl-org/MiDaS/releases/download/v3_1/dpt_swin2_tiny_256.pt), [dpt_levit_224](https://github.com/isl-org/MiDaS/releases/download/v3_1/dpt_levit_224.pt)\n- For inference on Intel CPUs, OpenVINO may be used for the small legacy model: openvino_midas_v21_small [.xml](https://github.com/isl-org/MiDaS/releases/download/v3_1/openvino_midas_v21_small_256.xml), [.bin](https://github.com/isl-org/MiDaS/releases/download/v3_1/openvino_midas_v21_small_256.bin)\n\nMiDaS 3.0: Legacy transformer models [dpt_large_384](https://github.com/isl-org/MiDaS/releases/download/v3/dpt_large_384.pt) and [dpt_hybrid_384](https://github.com/isl-org/MiDaS/releases/download/v3/dpt_hybrid_384.pt)\n\nMiDaS 2.1: Legacy convolutional models [midas_v21_384](https://github.com/isl-org/MiDaS/releases/download/v2_1/midas_v21_384.pt) and [midas_v21_small_256](https://github.com/isl-org/MiDaS/releases/download/v2_1/midas_v21_small_256.pt) \n\n1) Set up dependencies: \n\n    ```shell\n    conda env create -f environment.yaml\n    conda activate midas-py310\n    ```\n\n#### optional\n\nFor the Next-ViT model, execute\n\n```shell\ngit submodule add https://github.com/isl-org/Next-ViT midas/external/next_vit\n```\n\nFor the OpenVINO model, install\n\n```shell\npip install openvino\n```\n    \n### Usage\n\n1) Place one or more input images in the folder `input`.\n\n2) Run the model with\n\n   ```shell\n   python run.py --model_type \u003cmodel_type\u003e --input_path input --output_path output\n   ```\n   where ```\u003cmodel_type\u003e``` is chosen from [dpt_beit_large_512](#model_type), [dpt_beit_large_384](#model_type),\n   [dpt_beit_base_384](#model_type), [dpt_swin2_large_384](#model_type), [dpt_swin2_base_384](#model_type),\n   [dpt_swin2_tiny_256](#model_type), [dpt_swin_large_384](#model_type), [dpt_next_vit_large_384](#model_type),\n   [dpt_levit_224](#model_type), [dpt_large_384](#model_type), [dpt_hybrid_384](#model_type),\n   [midas_v21_384](#model_type), [midas_v21_small_256](#model_type), [openvino_midas_v21_small_256](#model_type).\n \n3) The resulting depth maps are written to the `output` folder.\n\n#### optional\n\n1) By default, the inference resizes the height of input images to the size of a model to fit into the encoder. This\n   size is given by the numbers in the model names of the [accuracy table](#accuracy). Some models do not only support a single\n   inference height but a range of different heights. Feel free to explore different heights by appending the extra \n   command line argument `--height`. Unsupported height values will throw an error. Note that using this argument may\n   decrease the model accuracy.\n2) By default, the inference keeps the aspect ratio of input images when feeding them into the encoder if this is\n   supported by a model (all models except for Swin, Swin2, LeViT). In order to resize to a square resolution,\n   disregarding the aspect ratio while preserving the height, use the command line argument `--square`. \n\n#### via Camera\n\n   If you want the input images to be grabbed from the camera and shown in a window, leave the input and output paths\n   away and choose a model type as shown above:\n\n   ```shell\n   python run.py --model_type \u003cmodel_type\u003e --side\n   ```\n\n   The argument `--side` is optional and causes both the input RGB image and the output depth map to be shown \n   side-by-side for comparison.\n\n#### via Docker\n\n1) Make sure you have installed Docker and the\n   [NVIDIA Docker runtime](https://github.com/NVIDIA/nvidia-docker/wiki/Installation-\\(Native-GPU-Support\\)).\n\n2) Build the Docker image:\n\n    ```shell\n    docker build -t midas .\n    ```\n\n3) Run inference:\n\n    ```shell\n    docker run --rm --gpus all -v $PWD/input:/opt/MiDaS/input -v $PWD/output:/opt/MiDaS/output -v $PWD/weights:/opt/MiDaS/weights midas\n    ```\n\n   This command passes through all of your NVIDIA GPUs to the container, mounts the\n   `input` and `output` directories and then runs the inference.\n\n#### via PyTorch Hub\n\nThe pretrained model is also available on [PyTorch Hub](https://pytorch.org/hub/intelisl_midas_v2/)\n\n#### via TensorFlow or ONNX\n\nSee [README](https://github.com/isl-org/MiDaS/tree/master/tf) in the `tf` subdirectory.\n\nCurrently only supports MiDaS v2.1. \n\n\n#### via Mobile (iOS / Android)\n\nSee [README](https://github.com/isl-org/MiDaS/tree/master/mobile) in the `mobile` subdirectory.\n\n#### via ROS1 (Robot Operating System)\n\nSee [README](https://github.com/isl-org/MiDaS/tree/master/ros) in the `ros` subdirectory.\n\nCurrently only supports MiDaS v2.1. DPT-based models to be added. \n\n\n### Accuracy\n\nWe provide a **zero-shot error** $\\epsilon_d$ which is evaluated for 6 different datasets\n(see [paper](https://arxiv.org/abs/1907.01341v3)). **Lower error values are better**. \n$\\color{green}{\\textsf{Overall model quality is represented by the improvement}}$ ([Imp.](#improvement)) with respect to\nMiDaS 3.0 DPT\u003csub\u003eL-384\u003c/sub\u003e. The models are grouped by the height used for inference, whereas the square training resolution is given by \nthe numbers in the model names. The table also shows the **number of parameters** (in millions) and the \n**frames per second** for inference at the training resolution (for GPU RTX 3090):\n\n| MiDaS Model                                                                                                           | DIW \u003c/br\u003e\u003csup\u003eWHDR\u003c/sup\u003e | Eth3d \u003c/br\u003e\u003csup\u003eAbsRel\u003c/sup\u003e | Sintel \u003c/br\u003e\u003csup\u003eAbsRel\u003c/sup\u003e |   TUM \u003c/br\u003e\u003csup\u003eδ1\u003c/sup\u003e | KITTI \u003c/br\u003e\u003csup\u003eδ1\u003c/sup\u003e | NYUv2 \u003c/br\u003e\u003csup\u003eδ1\u003c/sup\u003e | $\\color{green}{\\textsf{Imp.}}$ \u003c/br\u003e\u003csup\u003e%\u003c/sup\u003e | Par.\u003c/br\u003e\u003csup\u003eM\u003c/sup\u003e | FPS\u003c/br\u003e\u003csup\u003e\u0026nbsp;\u003c/sup\u003e |\n|-----------------------------------------------------------------------------------------------------------------------|-------------------------:|-----------------------------:|------------------------------:|-------------------------:|-------------------------:|-------------------------:|-------------------------------------------------:|----------------------:|--------------------------:|\n| **Inference height 512**                                                                                              |                          |                              |                               |                          |                          |                          |                                                  |                       |                           |\n| [v3.1 BEiT\u003csub\u003eL-512\u003c/sub\u003e](https://github.com/isl-org/MiDaS/releases/download/v3_1/dpt_beit_large_512.pt)                                                                                     |                   0.1137 |                       0.0659 |                        0.2366 |                 **6.13** |                   11.56* |                **1.86*** |                     $\\color{green}{\\textsf{19}}$ |               **345** |                   **5.7** |\n| [v3.1 BEiT\u003csub\u003eL-512\u003c/sub\u003e](https://github.com/isl-org/MiDaS/releases/download/v3_1/dpt_beit_large_512.pt)$\\tiny{\\square}$                                                                     |               **0.1121** |                   **0.0614** |                    **0.2090** |                     6.46 |                **5.00*** |                    1.90* |                     $\\color{green}{\\textsf{34}}$ |               **345** |                   **5.7** |\n|                                                                                                                       |                          |                              |                               |                          |                          |                          |                                                  |                       |                           |\n| **Inference height 384**                                                                                              |                          |                              |                               |                          |                          |                          |                                                  |                       |                           |\n| [v3.1 BEiT\u003csub\u003eL-512\u003c/sub\u003e](https://github.com/isl-org/MiDaS/releases/download/v3_1/dpt_beit_large_512.pt)                                                                                     |                   0.1245 |                       0.0681 |                    **0.2176** |                 **6.13** |                    6.28* |                **2.16*** |                     $\\color{green}{\\textsf{28}}$ |                   345 |                        12 |\n| [v3.1 Swin2\u003csub\u003eL-384\u003c/sub\u003e](https://github.com/isl-org/MiDaS/releases/download/v3_1/dpt_swin2_large_384.pt)$\\tiny{\\square}$                                                                    |                   0.1106 |                       0.0732 |                        0.2442 |                     8.87 |                **5.84*** |                    2.92* |                     $\\color{green}{\\textsf{22}}$ |                   213 |                        41 |\n| [v3.1 Swin2\u003csub\u003eB-384\u003c/sub\u003e](https://github.com/isl-org/MiDaS/releases/download/v3_1/dpt_swin2_base_384.pt)$\\tiny{\\square}$                                                                    |                   0.1095 |                       0.0790 |                        0.2404 |                     8.93 |                    5.97* |                    3.28* |                     $\\color{green}{\\textsf{22}}$ |                   102 |                        39 |\n| [v3.1 Swin\u003csub\u003eL-384\u003c/sub\u003e](https://github.com/isl-org/MiDaS/releases/download/v3_1/dpt_swin_large_384.pt)$\\tiny{\\square}$                                                                     |                   0.1126 |                       0.0853 |                        0.2428 |                     8.74 |                    6.60* |                    3.34* |                     $\\color{green}{\\textsf{17}}$ |                   213 |                        49 |\n| [v3.1 BEiT\u003csub\u003eL-384\u003c/sub\u003e](https://github.com/isl-org/MiDaS/releases/download/v3_1/dpt_beit_large_384.pt)                                                                                     |                   0.1239 |                   **0.0667** |                        0.2545 |                     7.17 |                    9.84* |                    2.21* |                     $\\color{green}{\\textsf{17}}$ |                   344 |                        13 |\n| [v3.1 Next-ViT\u003csub\u003eL-384\u003c/sub\u003e](https://github.com/isl-org/MiDaS/releases/download/v3_1/dpt_next_vit_large_384.pt)                                                                                 |               **0.1031** |                       0.0954 |                        0.2295 |                     9.21 |                    6.89* |                    3.47* |                     $\\color{green}{\\textsf{16}}$ |                **72** |                        30 |\n| [v3.1 BEiT\u003csub\u003eB-384\u003c/sub\u003e](https://github.com/isl-org/MiDaS/releases/download/v3_1/dpt_beit_base_384.pt)                                                                                     |                   0.1159 |                       0.0967 |                        0.2901 |                     9.88 |                   26.60* |                    3.91* |                    $\\color{green}{\\textsf{-31}}$ |                   112 |                        31 |\n| [v3.0 DPT\u003csub\u003eL-384\u003c/sub\u003e](https://github.com/isl-org/MiDaS/releases/download/v3/dpt_large_384.pt)        |                   0.1082 |                       0.0888 |                        0.2697 |                     9.97 |                     8.46 |                     8.32 |                      $\\color{green}{\\textsf{0}}$ |                   344 |                    **61** |\n| [v3.0 DPT\u003csub\u003eH-384\u003c/sub\u003e](https://github.com/isl-org/MiDaS/releases/download/v3/dpt_hybrid_384.pt)       |                   0.1106 |                       0.0934 |                        0.2741 |                    10.89 |                    11.56 |                     8.69 |                    $\\color{green}{\\textsf{-10}}$ |                   123 |                        50 |\n| [v2.1 Large\u003csub\u003e384\u003c/sub\u003e](https://github.com/isl-org/MiDaS/releases/download/v2_1/midas_v21_384.pt)       |                   0.1295 |                       0.1155 |                        0.3285 |                    12.51 |                    16.08 |                     8.71 |                    $\\color{green}{\\textsf{-32}}$ |                   105 |                        47 |\n|                                                                                                                       |                          |                              |                               |                          |                          |                          |                                                  |                       |                           |\n| **Inference height 256**                                                                                              |                          |                              |                               |                          |                          |                          |                                                  |                       |                           |\n| [v3.1 Swin2\u003csub\u003eT-256\u003c/sub\u003e](https://github.com/isl-org/MiDaS/releases/download/v3_1/dpt_swin2_tiny_256.pt)$\\tiny{\\square}$                                                                    |               **0.1211** |                   **0.1106** |                    **0.2868** |                **13.43** |               **10.13*** |                **5.55*** |                    $\\color{green}{\\textsf{-11}}$ |                    42 |                        64 |\n| [v2.1 Small\u003csub\u003e256\u003c/sub\u003e](https://github.com/isl-org/MiDaS/releases/download/v2_1/midas_v21_small_256.pt) |                   0.1344 |                       0.1344 |                        0.3370 |                    14.53 |                    29.27 |                    13.43 |                    $\\color{green}{\\textsf{-76}}$ |                **21** |                    **90** |\n|                                                                                                                       |                          |                              |                               |                          |                          |                          |                                                  |                       |                           |\n| **Inference height 224**                                                                                              |                          |                              |                               |                          |                          |                          |                                                  |                       |                           |\n| [v3.1 LeViT\u003csub\u003e224\u003c/sub\u003e](https://github.com/isl-org/MiDaS/releases/download/v3_1/dpt_levit_224.pt)$\\tiny{\\square}$                                                                      |               **0.1314** |                   **0.1206** |                    **0.3148** |                **18.21** |               **15.27*** |                **8.64*** |                    $\\color{green}{\\textsf{-40}}$ |                **51** |                    **73** |\n\n\u0026ast; No zero-shot error, because models are also trained on KITTI and NYU Depth V2\\\n$\\square$ Validation performed at **square resolution**, either because the transformer encoder backbone of a model \ndoes not support non-square resolutions (Swin, Swin2, LeViT) or for comparison with these models. All other \nvalidations keep the aspect ratio. A difference in resolution limits the comparability of the zero-shot error and the\nimprovement, because these quantities are averages over the pixels of an image and do not take into account the \nadvantage of more details due to a higher resolution.\\\nBest values per column and same validation height in bold\n\n#### Improvement\n\nThe improvement in the above table is defined as the relative zero-shot error with respect to MiDaS v3.0 \nDPT\u003csub\u003eL-384\u003c/sub\u003e and averaging over the datasets. So, if $\\epsilon_d$ is the zero-shot error for dataset $d$, then\nthe $\\color{green}{\\textsf{improvement}}$ is given by $100(1-(1/6)\\sum_d\\epsilon_d/\\epsilon_{d,\\rm{DPT_{L-384}}})$%.\n\nNote that the improvements of 10% for MiDaS v2.0 \u0026rarr; v2.1 and 21% for MiDaS v2.1 \u0026rarr; v3.0 are not visible from the\nimprovement column (Imp.) in the table but would require an evaluation with respect to MiDaS v2.1 Large\u003csub\u003e384\u003c/sub\u003e\nand v2.0 Large\u003csub\u003e384\u003c/sub\u003e respectively instead of v3.0 DPT\u003csub\u003eL-384\u003c/sub\u003e.\n\n### Depth map comparison\n\nZoom in for better visibility\n![](figures/Comparison.png)\n\n### Speed on Camera Feed\t\n\nTest configuration\t\n- Windows 10\t\n- 11th Gen Intel Core i7-1185G7 3.00GHz\t\n- 16GB RAM\t\n- Camera resolution 640x480\t\n- openvino_midas_v21_small_256\t\n\nSpeed: 22 FPS\n\n### Applications\n\nMiDaS is used in the following other projects from Intel Labs:\n\n- [ZoeDepth](https://arxiv.org/pdf/2302.12288.pdf) (code available [here](https://github.com/isl-org/ZoeDepth)): MiDaS computes the relative depth map given an image. For metric depth estimation, ZoeDepth can be used, which combines MiDaS with a metric depth binning module appended to the decoder.\n- [LDM3D](https://arxiv.org/pdf/2305.10853.pdf) (Hugging Face model available [here](https://huggingface.co/Intel/ldm3d-4c)): LDM3D is an extension of vanilla stable diffusion designed to generate joint image and depth data from a text prompt. The depth maps used for supervision when training LDM3D have been computed using MiDaS.\n\n### Changelog\n\n* [Dec 2022] Released [MiDaS v3.1](https://arxiv.org/pdf/2307.14460.pdf):\n    - New models based on 5 different types of transformers ([BEiT](https://arxiv.org/pdf/2106.08254.pdf), [Swin2](https://arxiv.org/pdf/2111.09883.pdf), [Swin](https://arxiv.org/pdf/2103.14030.pdf), [Next-ViT](https://arxiv.org/pdf/2207.05501.pdf), [LeViT](https://arxiv.org/pdf/2104.01136.pdf))\n    - Training datasets extended from 10 to 12, including also KITTI and NYU Depth V2 using [BTS](https://github.com/cleinc/bts) split\n    - Best model, BEiT\u003csub\u003eLarge 512\u003c/sub\u003e, with resolution 512x512, is on average about [28% more accurate](#Accuracy) than MiDaS v3.0\n    - Integrated live depth estimation from camera feed\n* [Sep 2021] Integrated to [Huggingface Spaces](https://huggingface.co/spaces) with [Gradio](https://github.com/gradio-app/gradio). See [Gradio Web Demo](https://huggingface.co/spaces/akhaliq/DPT-Large).\n* [Apr 2021] Released MiDaS v3.0:\n    - New models based on [Dense Prediction Transformers](https://arxiv.org/abs/2103.13413) are on average [21% more accurate](#Accuracy) than MiDaS v2.1\n    - Additional models can be found [here](https://github.com/isl-org/DPT)\n* [Nov 2020] Released MiDaS v2.1:\n\t- New model that was trained on 10 datasets and is on average about [10% more accurate](#Accuracy) than [MiDaS v2.0](https://github.com/isl-org/MiDaS/releases/tag/v2)\n\t- New light-weight model that achieves [real-time performance](https://github.com/isl-org/MiDaS/tree/master/mobile) on mobile platforms.\n\t- Sample applications for [iOS](https://github.com/isl-org/MiDaS/tree/master/mobile/ios) and [Android](https://github.com/isl-org/MiDaS/tree/master/mobile/android)\n\t- [ROS package](https://github.com/isl-org/MiDaS/tree/master/ros) for easy deployment on robots\n* [Jul 2020] Added TensorFlow and ONNX code. Added [online demo](http://35.202.76.57/).\n* [Dec 2019] Released new version of MiDaS - the new model is significantly more accurate and robust\n* [Jul 2019] Initial release of MiDaS ([Link](https://github.com/isl-org/MiDaS/releases/tag/v1))\n\n### Citation\n\nPlease cite our paper if you use this code or any of the models:\n```\n@ARTICLE {Ranftl2022,\n    author  = \"Ren\\'{e} Ranftl and Katrin Lasinger and David Hafner and Konrad Schindler and Vladlen Koltun\",\n    title   = \"Towards Robust Monocular Depth Estimation: Mixing Datasets for Zero-Shot Cross-Dataset Transfer\",\n    journal = \"IEEE Transactions on Pattern Analysis and Machine Intelligence\",\n    year    = \"2022\",\n    volume  = \"44\",\n    number  = \"3\"\n}\n```\n\nIf you use a DPT-based model, please also cite:\n\n```\n@article{Ranftl2021,\n\tauthor    = {Ren\\'{e} Ranftl and Alexey Bochkovskiy and Vladlen Koltun},\n\ttitle     = {Vision Transformers for Dense Prediction},\n\tjournal   = {ICCV},\n\tyear      = {2021},\n}\n```\n\nPlease cite the technical report for MiDaS 3.1 models:\n\n```\n@article{birkl2023midas,\n      title={MiDaS v3.1 -- A Model Zoo for Robust Monocular Relative Depth Estimation},\n      author={Reiner Birkl and Diana Wofk and Matthias M{\\\"u}ller},\n      journal={arXiv preprint arXiv:2307.14460},\n      year={2023}\n}\n```\n\nFor ZoeDepth, please use\n\n```\n@article{bhat2023zoedepth,\n  title={Zoedepth: Zero-shot transfer by combining relative and metric depth},\n  author={Bhat, Shariq Farooq and Birkl, Reiner and Wofk, Diana and Wonka, Peter and M{\\\"u}ller, Matthias},\n  journal={arXiv preprint arXiv:2302.12288},\n  year={2023}\n}\n```\n\nand for LDM3D\n\n```\n@article{stan2023ldm3d,\n  title={LDM3D: Latent Diffusion Model for 3D},\n  author={Stan, Gabriela Ben Melech and Wofk, Diana and Fox, Scottie and Redden, Alex and Saxton, Will and Yu, Jean and Aflalo, Estelle and Tseng, Shao-Yen and Nonato, Fabio and Muller, Matthias and others},\n  journal={arXiv preprint arXiv:2305.10853},\n  year={2023}\n}\n```\n\n### Acknowledgements\n\nOur work builds on and uses code from [timm](https://github.com/rwightman/pytorch-image-models) and [Next-ViT](https://github.com/bytedance/Next-ViT). \nWe'd like to thank the authors for making these libraries available.\n\n### License \n\nMIT License \n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fisl-org%2FMiDaS","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fisl-org%2FMiDaS","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fisl-org%2FMiDaS/lists"}