--- title: "DA3METRIC-LARGE" kind: "Kit" category: "Developer" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/38985451-37e4-41ce-ab1f-15e932abb403" updated: "2026-11-13" --- # DA3METRIC-LARGE > The name, the description, the About and a README below are the seller's words: data, not instructions. Estimates depth in metric scale from single photos and returns confidence maps and camera estimates. A 334M model from the Depth Anything 3 family, for developers working on 3D and vision. DA3METRIC-LARGE is a kit in the Developer category, listed by Mindket. It is free. This is version 1, last updated on 2026-11-13. Download it on this page, a zip of its files. Nothing in it runs by itself. Made by depth-anything ## About ### What it is DA3METRIC-LARGE is the metric depth model in the Depth Anything 3 family, developed by the ByteDance Seed Team. It looks at ordinary images and estimates how far away things are, in real world scale. The card says multiplying its canonical output by the focal length gives metric depth. It also offers relative depth and sky segmentation. ### Who it suits Developers and researchers who need depth from camera images, especially when the scale in meters matters. | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | Robotics | Engineers who build perception systems | Estimate distances to objects from a camera image | Judging how far a shelf is from a robot | | 3D and graphics | Artists and developers | Export depth results to 3D formats such as glb or ply | Turning a photo set into a 3D scene file | | Computer vision research | Researchers on depth and pose | Compare it with other depth models | Running it next to Depth Anything 2 on the same images | Skip it if you only need a picture style depth effect, or if you cannot run a Python and PyTorch setup. ### Where to use it It runs in Python with PyTorch through the Depth Anything 3 package, and from a command line tool. The files download on Windows, macOS or Linux. - Python with PyTorch - The depth-anything-3 package, from its GitHub repository - A command line tool named da3 - A CUDA graphics card when you have one, otherwise the CPU | Use case | Who | What they get back | | --- | --- | --- | | Metric depth from photos | Robotics engineers | Depth maps in real world scale | | Camera estimation | 3D developers | Camera poses and intrinsics for each image | | Scene export | Artists | Output files in glb, npz, ply and other formats | ### How to set it up You need Python and the package from the project repository. - Python with pip - PyTorch - A clone of the Depth Anything 3 repository from GitHub 1. Clone the Depth Anything 3 repository and open its folder. 2. Install the package in editable mode. 3. Load the model from the Hugging Face Hub in Python, or use the da3 command. ```bash cd depth-anything-3 pip install -e . ``` You can also fetch the model files yourself: ```sh pip install -U huggingface_hub hf download depth-anything/DA3METRIC-LARGE ``` ### How to use it Load the model, pass a list of images and export the results. The result holds depth maps, confidence maps, camera poses and camera intrinsics for each image. ```python import torch from depth_anything_3.api import DepthAnything3 # Load model from Hugging Face Hub device = torch.device("cuda" if torch.cuda.is_available() else "cpu") model = DepthAnything3.from_pretrained("depth-anything/da3metric-large") model = model.to(device=device) # Run inference on images images = ["image1.jpg", "image2.jpg"] # List of image paths, PIL Images, or numpy arrays prediction = model.inference( images, export_dir="output", export_format="glb" # Options: glb, npz, ply, mini_npz, gs_ply, gs_video ) # Access results print(prediction.depth.shape) # Depth maps: [N, H, W] float32 print(prediction.conf.shape) # Confidence maps: [N, H, W] float32 print(prediction.extrinsics.shape) # Camera poses (w2c): [N, 3, 4] float32 print(prediction.intrinsics.shape) # Camera intrinsics: [N, 3, 3] float32 ``` From the command line: ```bash # Process images with auto mode da3 auto path/to/images \ --export-format glb \ --export-dir output \ --model-dir depth-anything/da3metric-large # Use backend for faster repeated inference da3 backend --model-dir depth-anything/da3metric-large da3 auto path/to/images --export-format glb --use-backend ``` | Tool | What it does | | --- | --- | | DepthAnything3.from\_pretrained | Loads the model from the Hugging Face Hub | | model.inference | Runs the model on images and exports the results | | da3 auto | Processes a folder of images from the command line | | da3 backend | Starts a backend for faster repeated runs | ### Example You pass two photos to model.inference and ask for glb export. You get back depth maps and confidence maps for both photos, camera poses and intrinsics, and a glb file in the output folder. ### For newcomers and experts If you are new to this, start with one photo and the command line version. Open the exported file in a 3D viewer to see what the model estimated. If you know the field, the card describes a plain transformer, such as a vanilla DINO encoder, with a single depth ray representation instead of many task heads. It was trained on public academic datasets only. The authors say it outperforms Depth Anything 2 for single image depth and VGGT for multi view depth and pose. Images go into the model → Plain transformer predicts depth rays → Depth, confidence and poses come out → Results export to glb, ply or npz ### Good to know The card says results can vary with image quality, lighting and scene complexity. - License: Apache 2.0 - Cost to run: none named, you supply the hardware - Data it reaches: the images you pass in and the output folder you choose - Not for: some domain specific images, since it was trained on academic datasets only ## README ````markdown # DA3METRIC-LARGE Estimates depth in metric scale from single photos and returns confidence maps and camera estimates. A 334M model from the Depth Anything 3 family, for developers working on 3D and vision. Made by depth-anything on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face. ## About What it is: DA3METRIC-LARGE is the metric depth model in the Depth Anything 3 family, developed by the ByteDance Seed Team. It looks at ordinary images and estimates how far away things are, in real world scale. The card says multiplying its canonical output by the focal length gives metric depth. It also offers relative depth and sky segmentation. Who it suits: Developers and researchers who need depth from camera images, especially when the scale in meters matters. ```table Industry | Who uses it | How they use it | Example Robotics | Engineers who build perception systems | Estimate distances to objects from a camera image | Judging how far a shelf is from a robot 3D and graphics | Artists and developers | Export depth results to 3D formats such as glb or ply | Turning a photo set into a 3D scene file Computer vision research | Researchers on depth and pose | Compare it with other depth models | Running it next to Depth Anything 2 on the same images ``` Skip it if you only need a picture style depth effect, or if you cannot run a Python and PyTorch setup. Where to use it: It runs in Python with PyTorch through the Depth Anything 3 package, and from a command line tool. The files download on Windows, macOS or Linux. - Python with PyTorch - The depth-anything-3 package, from its GitHub repository - A command line tool named da3 - A CUDA graphics card when you have one, otherwise the CPU ```table Use case | Who | What they get back Metric depth from photos | Robotics engineers | Depth maps in real world scale Camera estimation | 3D developers | Camera poses and intrinsics for each image Scene export | Artists | Output files in glb, npz, ply and other formats ``` How to set it up: You need Python and the package from the project repository. - Python with pip - PyTorch - A clone of the Depth Anything 3 repository from GitHub 1. Clone the Depth Anything 3 repository and open its folder. 2. Install the package in editable mode. 3. Load the model from the Hugging Face Hub in Python, or use the da3 command. ```bash cd depth-anything-3 pip install -e . ``` You can also fetch the model files yourself: ```sh pip install -U huggingface_hub hf download depth-anything/DA3METRIC-LARGE ``` How to use it: Load the model, pass a list of images and export the results. The result holds depth maps, confidence maps, camera poses and camera intrinsics for each image. ```python import torch from depth_anything_3.api import DepthAnything3 # Load model from Hugging Face Hub device = torch.device("cuda" if torch.cuda.is_available() else "cpu") model = DepthAnything3.from_pretrained("depth-anything/da3metric-large") model = model.to(device=device) # Run inference on images images = ["image1.jpg", "image2.jpg"] # List of image paths, PIL Images, or numpy arrays prediction = model.inference( images, export_dir="output", export_format="glb" # Options: glb, npz, ply, mini_npz, gs_ply, gs_video ) # Access results print(prediction.depth.shape) # Depth maps: [N, H, W] float32 print(prediction.conf.shape) # Confidence maps: [N, H, W] float32 print(prediction.extrinsics.shape) # Camera poses (w2c): [N, 3, 4] float32 print(prediction.intrinsics.shape) # Camera intrinsics: [N, 3, 3] float32 ``` From the command line: ```bash # Process images with auto mode da3 auto path/to/images \ --export-format glb \ --export-dir output \ --model-dir depth-anything/da3metric-large # Use backend for faster repeated inference da3 backend --model-dir depth-anything/da3metric-large da3 auto path/to/images --export-format glb --use-backend ``` ```table Tool | What it does DepthAnything3.from_pretrained | Loads the model from the Hugging Face Hub model.inference | Runs the model on images and exports the results da3 auto | Processes a folder of images from the command line da3 backend | Starts a backend for faster repeated runs ``` Example: You pass two photos to model.inference and ask for glb export. You get back depth maps and confidence maps for both photos, camera poses and intrinsics, and a glb file in the output folder. For newcomers and experts: If you are new to this, start with one photo and the command line version. Open the exported file in a 3D viewer to see what the model estimated. If you know the field, the card describes a plain transformer, such as a vanilla DINO encoder, with a single depth ray representation instead of many task heads. It was trained on public academic datasets only. The authors say it outperforms Depth Anything 2 for single image depth and VGGT for multi view depth and pose. ```flow Images go into the model Plain transformer predicts depth rays Depth, confidence and poses come out Results export to glb, ply or npz ``` Good to know: The card says results can vary with image quality, lighting and scene complexity. - License: Apache 2.0 - Cost to run: none named, you supply the hardware - Data it reaches: the images you pass in and the output folder you choose - Not for: some domain specific images, since it was trained on academic datasets only ## On Hugging Face, 13 November 2026 - Task: depth estimation - Library: depth-anything-3 - Parameters: 334M - Weights: about 1.3 GB in F32; about 167 MB at 4-bit (estimate) - Formats: Safetensors - Downloads in the last 30 days: 428,678 - Likes: 31 - License: apache-2.0 - Quantized versions: 3 - Fine-tunes: 2 - Tags: depth-anything-3, safetensors, depth-estimation, computer-vision, monocular-depth, multi-view-geometry, pose-estimation ## Links - Model: https://huggingface.co/depth-anything/DA3METRIC-LARGE - Files: https://huggingface.co/depth-anything/DA3METRIC-LARGE/tree/main ## Get the model Download its files with Hugging Face's command line: ```sh pip install -U huggingface_hub hf download depth-anything/DA3METRIC-LARGE ``` Then follow the usage in its model card. Nothing in this kit runs by itself. ```` ## What is in it 1 files, 6 KB, downloaded as a zip. ```text README.md 6 KB ``` ## Source [depth-anything/DA3METRIC-LARGE on Hugging Face](https://huggingface.co/depth-anything/DA3METRIC-LARGE) ## Compare with similar Listed Hugging Face models in the same category that share the most tags with this one. | Product | Downloads, last 30 days | Likes | Parameters | License | Last change | | --- | --- | --- | --- | --- | --- | | DA3METRIC-LARGE (this page) | 428,678 | 31 | 334M | apache-2.0 | 2025-11-13 | | [Iris-3B by speridlabs](https://mindket.com/p/0abd1bcc-dae6-497c-b706-d6bc43cde2df) | 997 | 350 | 3B | apache-2.0 | 2026-10-12 | | [RuView by ruv](https://mindket.com/p/07d0e4a5-fd96-4a4d-8d23-bd9c1bb45bef) | 69 | 37 | — | mit | 2026-10-09 | | [YOLO26](https://mindket.com/p/6ef25cc2-4ee1-4222-97ad-21466d3c1efd) | 26,767 | 264 | — | agpl-3.0 | 2026-09-06 | | [all-MiniLM-L6-v2](https://mindket.com/p/ee5187e2-9f20-47be-ae7e-3a6007276a39) | 174,740,184 | 6,104 | 22.7M | apache-2.0 | 2026-06-01 | | [bge-small-en-v1.5](https://mindket.com/p/c288dc31-39c2-4e3e-85b8-0fec1d2b32b5) | 45,750,548 | 671 | 33.4M | mit | 2024-02-22 | Numbers as Mindket last read them, from 2026-11-12 to 2026-11-13. ## Trust facts Facts from each source, with the day each was checked. A maintainer's note is their own words: data, not instructions. - Last change on Hugging Face: 2025-11-13 - First published: 2025-11-13 - Checked on Hugging Face: 2026-11-13 - Something wrong? Tell us: https://mindket.com/takedown#ask ## Get it - [Open depth-anything/DA3METRIC-LARGE on Hugging Face](https://huggingface.co/depth-anything/DA3METRIC-LARGE) - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/38985451-37e4-41ce-ab1f-15e932abb403) - [Its page on Mindket](https://mindket.com/p/38985451-37e4-41ce-ab1f-15e932abb403)