Book a demoTry for free
EnterprisePricingBlogDocsLog inBook a demo

Train models on your data.
Deploy them on your GPUs.

Train computer vision models such as YOLO26, DEIMv2 and RT-DETRv2 on your labeled data without code, then deploy them on your own GPUs to pre-label data or serve predictions through an API.

Book a demoTry for free

Train state-of-the-art models on your data, without code.

Pick a model, the classes and the split, and start. Training runs on your GPU and the metrics update live.

  • Pretrained start

    From a checkpoint pretrained on COCO, or from a model you trained before.

  • Live metrics

    Losses, mAP and logs update every epoch, with TensorBoard one click away.

  • Evaluation and export

    Runs end with an evaluation report by default, and detectors export to ONNX or TensorRT.

Train YOLO26 step by stepPrepare the training data
  • Detection · Segmentation · Pose

    YOLO26 to YOLOv8

    YOLO26, YOLO12, YOLO11 and earlier checkpoints.

  • Detection

    DEIMv2

    Real-time detection built on DINOv3 features.

  • Detection

    RT-DETRv2

    A real-time detection transformer.

  • Promptable segmentation

    Segment Anything

    Fine-tuned on your objects for images and videos.

  • Instance segmentation

    MaskDINO

    Detection and segmentation in one transformer.

  • Detection · Instance segmentation

    MMDetection 3.0

    Dozens of architectures from OpenMMLab.

  • Semantic segmentation

    MMSegmentation

    Pixel-level models from OpenMMLab.

  • Semantic segmentation

    UNet

    A light, proven baseline for every pixel.

  • Classification

    MMPretrain

    Image classifiers from OpenMMLab.

  • 3D detection

    MMDetection3D

    Object detection in point clouds.

  • And every other model in the EcosystemNew architectures arrive as apps, ready to train and serve on your GPUs.

Every run is an experiment you can compare and reproduce.

One table keeps each model with its data version, settings, checkpoints and evaluation report. From any row, deploy the model, apply it to new data or train again with the same settings.

Evaluate and compare modelsExperiments in the docs

Run training and inference on your own GPUs.

Connect a workstation, a server or a cloud machine once, then start training, serving and prediction on it from the browser.

  • One command to connect

    The agent is a small Docker container. It needs no static IP address or network setup.

  • Your hardware stays yours

    Only your team can use your agents, and only the tasks you start run on them.

  • Many machines, side by side

    Install agents on Linux, Windows, cloud machines or Kubernetes and run experiments in parallel, each task in its own container.

Try for freeThe free Community plan connects one GPU machine.
How to connect agents on Linux, Windows and Kubernetes

Deploy a model in a few clicks or with one API call.

Serve a checkpoint on your agent from the browser, or start it from Python and send it images, videos or whole projects.

  • PyTorch, ONNX Runtime or TensorRT
  • On your own server or in Docker, outside the platform
  • Checkpoints download in open formats, so there is no lock-in
Prediction APIUse models outside Supervisely
import supervisely as sly

api = sly.Api.from_env()

# Deploy your checkpoint from Team Files on one of your agents
model = api.nn.deploy(
    model="/experiments/27_Lemons/392_RT-DETRv2/checkpoints/best.pth",
    device="cuda:0",
)

# Predict a folder of images, a video, a URL or an array
for p in model.predict(input="images/", conf=0.5):
    print(p.name, p.classes, p.scores)

Pre-label new data with the models you trained.

A deployed model labels whole projects, or the image an annotator has open. People review and fix its labels instead of drawing every object.

  • Whole projects and datasets

    Predictions merge with existing labels or go to a new project. On videos, tracking links boxes into objects.

  • Inside the labeling tools

    Annotators run the model on the image they have open and correct what it missed.

  • Live Training

    A model trains while the team labels and pre-labels each next image.

AI-assisted labeling

Frequently asked questions.

How do I train YOLO on a custom dataset in Supervisely?

Import or label your images and connect a GPU machine as an agent. Then start the YOLO training app on the project, pick a YOLO26 or older checkpoint, the classes and the train and validation split, and press Start. Checkpoints, logs and the evaluation report are saved with the experiment. Follow the YOLO26 tutorial for every step.

Can I train an object detection model without writing code?

Yes. You choose the model, classes, split and hyperparameters in a form and watch training live. You only need code to automate training through the API.

How do I deploy a trained model as an API?

Serve the checkpoint from the browser, or deploy it with one call from the Python SDK. The model runs as an inference server on your agent and returns classes, scores, boxes and masks for images, folders, videos or projects. Read the Prediction API guide.

Can I deploy a model on my own GPU or server?

Yes. Connect the machine to Supervisely as an agent with one command, on Linux, Windows through WSL, an AWS AMI or Kubernetes, and models are served on it. Without the platform, the same serving code runs on your server or in a Docker container, and checkpoints download in PyTorch, ONNX or TensorRT format. See how to connect agents.

Is model training available on the free plan?

Yes. Training, deployment, pre-labeling and checkpoint export are part of every plan, including the free Community plan, which connects one GPU machine. Pro connects two machines and adds inference APIs and active learning; Enterprise connects a cluster. Compare plans.