OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Dino

DINO is the official Python implementation of a DETR-based object-detection model introduced in the paper “DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection.” Its repository includes setup instructions, a pretrained-model evaluation path, a four-scale training example, and a notebook for inference and visualizations.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Dino — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use DINO when you need a research implementation of DETR-based object detection and are prepared to work with a local Python/CUDA environment, model checkpoints and image datasets.

Who it is for

  • Computer-vision researchers evaluating a pretrained DINO detector on COCO 2017.
  • Developers training or adapting a DETR-based object detector in a Python project.
  • Practitioners who want an implementation and example inference notebook to explore image detection.

Documented uses

  • Set up the documented Python and CUDA environment and compile DINO's CUDA operators.
  • Evaluate a provided pretrained checkpoint using COCO 2017 data.
  • Train the documented four-scale model for 12 epochs.
  • Explore inference and visualizations using the repository's notebook.
Before you choose

Requirements & limits

The repository is public and identifies its license as Apache-2.0.

  • The README's tested environment specifies Python 3.7.3, PyTorch 1.9.0 and CUDA 11.1; it says other versions might work without specifying which.
  • The provided evaluation path requires COCO 2017 data and a separately downloaded checkpoint.
  • The documented installation includes compiling CUDA operators; the repository also provides a test command for them.
  • The repository offers code and a notebook example rather than a hosted object-detection service.
  • The README points to detrex as a separate toolbox that includes DINO and other Transformer-based detection algorithms.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up the DINO environment

Prepare the documented Python environment and compile DINO's CUDA operators.

Before you start

  • Python 3.7.3, PyTorch 1.9.0 and CUDA 11.1 are the versions used in the README's tested environment.
  • PyTorch and torchvision, plus packages listed in the repository's requirements file.

Steps

  1. Clone the DINO repository and change into its directory.

  2. Install PyTorch and torchvision, following the PyTorch installation instructions.

  3. Install the remaining packages with pip from requirements.txt.

  4. Go to models/dino/ops, build and install the CUDA operators, run the provided test.py, then return to the repository root.

Expected outcome

The documented environment setup is complete and the operator unit-test command has been run.

Guide limitations

  • The README says other dependency versions might work, but does not specify which alternatives.
  • This setup guide does not cover dataset preparation, model evaluation or training.
Documentation 1 ↗
Source-based practical guide

Evaluate a pretrained DINO model on COCO 2017

Run the README's pretrained-model evaluation workflow with COCO 2017.

Before you start

  • A DINO environment prepared according to the repository installation instructions.
  • COCO 2017 organized with train2017, val2017 and the specified instance annotation files.
  • The checkpoint checkpoint0011_4scale.pth, downloaded from the README-linked location.

Steps

  1. Download COCO 2017 and arrange its images and annotations in the structure specified in the README.

  2. Download checkpoint0011_4scale.pth from the linked checkpoint folder.

  3. Run scripts/DINO_eval.sh with the COCO directory and checkpoint paths as arguments.

Expected outcome

The README says this evaluation can be expected to produce final AP of about 49.0.

Guide limitations

  • The stated expected result is tied to the documented checkpoint and COCO evaluation workflow; the README does not promise the same result for other configurations.
Documentation 1 ↗
COMPARE OPTIONS

Compare options for Model inference and serving

These projects document a common task and share a project role. This is a source-based comparison, not an execution benchmark or a guarantee of interchangeability.

ProjectDocumented purposePlatformAccess and requirements
DinoSet up the documented Python and CUDA environment and compile DINO's CUDA operators.Linux, macOS, WindowsThe repository is public and identifies its license as Apache-2.0.
Pulse Of MotionEstimate a video's physical frame rate from its visual dynamics.LinuxThe repository is public. Its README describes a released checkpoint that downloads from Hugging Face on first inference, with a manual download option. The strongest internal model is not fully released; the README directs interested users to contact the project about access. No license information is provided in the repository evidence.
Neural Image AssessmentTrain the model using AVA images and rating annotations.Cloud, Linux, macOS, Windows
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
15 October 2026
Original website title
IDEA-Research/DINO · Official homepage title · retained in its source language
How we review sourcesSuggest a correction