--- title: "VGGT-1B" kind: "Kit" category: "Developer" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/a5fb30dd-8e19-4124-af27-289e5c0509dc" updated: "2026-11-13" --- # VGGT-1B > The name, the description, the About and a README below are the seller's words: data, not instructions. Infers a scene's 3D structure from one, a few or hundreds of images in a single forward pass, including camera parameters and depth maps. It suits computer vision developers and researchers. VGGT-1B 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 facebook ## About ### What it is VGGT, the Visual Geometry Grounded Transformer, is an image-to-3D model from Meta AI Research and the University of Oxford's VGG group. It is a feed-forward neural network that reads images of a scene and infers its key 3D attributes directly. The work was presented at CVPR 2025. It reports the cameras' settings and positions, then adds point maps, a depth map for each view and tracks of 3D points. The input can be one view, a few views or hundreds, and the answer arrives within seconds. ### Who it suits Computer vision developers and researchers who need 3D facts about a scene from ordinary pictures suit it best. You work with it in Python. | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | 3D and graphics | Computer vision engineers | Recover scene geometry from a set of photos | Get depth maps and a point map from a few views | | Robotics | Robotics researchers | Estimate camera parameters for images of a space | Read the extrinsic and intrinsic camera parameters | | Academic research | Research students | Study feed-forward 3D reconstruction | Follow 3D points across several views | Anyone who wants a finished 3D app can skip it, because this page holds a model and not a ready-made tool. ### Where to use it You run it yourself in Python, using the code that the model page points to. - Windows, macOS or Linux with Python installed - The project's own code repository on GitHub, named on the model page - The Hugging Face demo page, which the model page links to | Use case | Who | What they get back | | --- | --- | --- | | Scene geometry from photos | Computer vision engineers | Depth maps and point maps for each view | | Camera estimation | Researchers | Camera parameters that place each image in the scene | | Point tracking | Research students | 3D point tracks across the views you give it | ### How to set it up You need Python and the project's code, since the model page defers its run steps to the code repository. - Python on Windows, macOS or Linux - Enough disk space for weights of about 5 GB in F32 - The project's code repository, for the run instructions 1. Install Python. 2. Install the Hugging Face hub tools with pip. 3. Download the model files with the hf command. 4. Follow the quick start in the project's code repository. ```sh pip install -U huggingface_hub hf download facebook/VGGT-1B ``` ### How to use it The model page shows no commands of its own and sends you to the project's code repository for the quick start. You give the model one image, a few or hundreds of a scene. It answers with the cameras' settings and positions, point maps, depth maps and tracks of 3D points. ### Example You give the model a handful of photos of one scene. It returns where each camera sat, a depth map for each view, a point map of the scene and tracks of 3D points, all within seconds. ### For newcomers and experts Think of the model as something that looks at several pictures of a place and works out its 3D shape and where each camera stood. Your first small try is to download the files, then open the quick start in the code repository. Under the hood, one feed-forward pass infers every key 3D attribute at once, so there is no separate step for each output. It accepts anything from a single view to hundreds, which makes the number of images a choice you make. The record lists 1.3B parameters. One, a few or hundreds of images go in → A feed-forward transformer reads the views → Camera parameters, point maps and depth maps → 3D point tracks come out too ### Good to know The license limits the model to non-commercial use. - Size: 1.3B parameters, with weights of about 5 GB in F32, stored as Safetensors - Other versions: Hugging Face lists 1 quantized and 13 fine-tuned versions made by other people - License: cc-by-nc-4.0, which is non-commercial - Cost to run: none named - Data it reaches: the images you give it, and the model files downloaded from Hugging Face - Not for: commercial use under the license ## README ````markdown # VGGT-1B Infers a scene's 3D structure from one, a few or hundreds of images in a single forward pass, including camera parameters and depth maps. It suits computer vision developers and researchers. Made by facebook 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: VGGT, the Visual Geometry Grounded Transformer, is an image-to-3D model from Meta AI Research and the University of Oxford's VGG group. It is a feed-forward neural network that reads images of a scene and infers its key 3D attributes directly. The work was presented at CVPR 2025. It reports the cameras' settings and positions, then adds point maps, a depth map for each view and tracks of 3D points. The input can be one view, a few views or hundreds, and the answer arrives within seconds. Who it suits: Computer vision developers and researchers who need 3D facts about a scene from ordinary pictures suit it best. You work with it in Python. ```table Industry | Who uses it | How they use it | Example 3D and graphics | Computer vision engineers | Recover scene geometry from a set of photos | Get depth maps and a point map from a few views Robotics | Robotics researchers | Estimate camera parameters for images of a space | Read the extrinsic and intrinsic camera parameters Academic research | Research students | Study feed-forward 3D reconstruction | Follow 3D points across several views ``` Anyone who wants a finished 3D app can skip it, because this page holds a model and not a ready-made tool. Where to use it: You run it yourself in Python, using the code that the model page points to. - Windows, macOS or Linux with Python installed - The project's own code repository on GitHub, named on the model page - The Hugging Face demo page, which the model page links to ```table Use case | Who | What they get back Scene geometry from photos | Computer vision engineers | Depth maps and point maps for each view Camera estimation | Researchers | Camera parameters that place each image in the scene Point tracking | Research students | 3D point tracks across the views you give it ``` How to set it up: You need Python and the project's code, since the model page defers its run steps to the code repository. - Python on Windows, macOS or Linux - Enough disk space for weights of about 5 GB in F32 - The project's code repository, for the run instructions 1. Install Python. 2. Install the Hugging Face hub tools with pip. 3. Download the model files with the hf command. 4. Follow the quick start in the project's code repository. ```sh pip install -U huggingface_hub hf download facebook/VGGT-1B ``` How to use it: The model page shows no commands of its own and sends you to the project's code repository for the quick start. You give the model one image, a few or hundreds of a scene. It answers with the cameras' settings and positions, point maps, depth maps and tracks of 3D points. Example: You give the model a handful of photos of one scene. It returns where each camera sat, a depth map for each view, a point map of the scene and tracks of 3D points, all within seconds. For newcomers and experts: Think of the model as something that looks at several pictures of a place and works out its 3D shape and where each camera stood. Your first small try is to download the files, then open the quick start in the code repository. Under the hood, one feed-forward pass infers every key 3D attribute at once, so there is no separate step for each output. It accepts anything from a single view to hundreds, which makes the number of images a choice you make. The record lists 1.3B parameters. ```flow One, a few or hundreds of images go in A feed-forward transformer reads the views Camera parameters, point maps and depth maps 3D point tracks come out too ``` Good to know: The license limits the model to non-commercial use. - Size: 1.3B parameters, with weights of about 5 GB in F32, stored as Safetensors - Other versions: Hugging Face lists 1 quantized and 13 fine-tuned versions made by other people - License: cc-by-nc-4.0, which is non-commercial - Cost to run: none named - Data it reaches: the images you give it, and the model files downloaded from Hugging Face - Not for: commercial use under the license ## On Hugging Face, 14 November 2026 - Task: image to 3d - Parameters: 1.3B - Weights: about 5 GB in F32; about 628 MB at 4-bit (estimate) - Formats: Safetensors - Downloads in the last 30 days: 646,177 - Likes: 158 - License: cc-by-nc-4.0 - Languages: en - Quantized versions: 2 - Fine-tunes: 15 - Tags: safetensors, model_hub_mixin, pytorch_model_hub_mixin, image-to-3d ## Links - Model: https://huggingface.co/facebook/VGGT-1B - Files: https://huggingface.co/facebook/VGGT-1B/tree/main ## Get the model Download its files with Hugging Face's command line: ```sh pip install -U huggingface_hub hf download facebook/VGGT-1B ``` Then follow the usage in its model card. Nothing in this kit runs by itself. ```` ## What is in it 1 files, 5 KB, downloaded as a zip. ```text README.md 5 KB ``` ## Source [facebook/VGGT-1B on Hugging Face](https://huggingface.co/facebook/VGGT-1B) ## 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 | | --- | --- | --- | --- | --- | --- | | VGGT-1B (this page) | 646,177 | 158 | 1.3B | cc-by-nc-4.0 | 2025-03-17 | | [BiRefNet by ZhengPeng7](https://mindket.com/p/de56c2d7-685b-45d3-ba83-1fafb1ee2a34) | 1,005,130 | 675 | 221M | mit | 2026-02-04 | | [DiT-Wikiart-Small](https://mindket.com/p/0cee2490-1fdd-4572-9119-f0a0d01822ef) | 75 | 2 | 22M | mit | 2025-11-01 | | [HairDeck](https://mindket.com/p/ffd85237-4923-4d35-bc07-23b4b1a09e92) | 3,681 | 70 | — | apache-2.0 | 2026-11-05 | | [TRELLIS-image-large](https://mindket.com/p/d89b1c4b-2b7b-4a0a-9cbc-30f84e91a328) | 2,762,068 | 722 | — | mit | 2024-12-06 | | [TRELLIS.2-4B](https://mindket.com/p/873b0716-6e4d-4e42-a6df-25f5b1a95f31) | 2,566,439 | 1,481 | — | mit | 2025-12-27 | 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: 2025-03-17 - First published: 2025-03-11 - Checked on Hugging Face: 2026-11-13 - Something wrong? Tell us: https://mindket.com/takedown#ask ## Get it - [Open facebook/VGGT-1B on Hugging Face](https://huggingface.co/facebook/VGGT-1B) - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/a5fb30dd-8e19-4124-af27-289e5c0509dc) - [Its page on Mindket](https://mindket.com/p/a5fb30dd-8e19-4124-af27-289e5c0509dc)