Ultralytics YOLO - Weights & Biases Documentation

Ultralytics provides computer vision models for tasks like image classification, object detection, image segmentation, and pose estimation. It hosts YOLOv8, an iteration in the YOLO series of real-time object detection models, along with other computer vision models such as SAM (Segment Anything Model), RT-DETR, and YOLO-NAS. Ultralytics also provides ready-to-use workflows for training, fine-tuning, and applying these models through an API. This page shows you how to integrate W&B with Ultralytics so that W&B automatically tracks and visualizes experiment metrics, model checkpoints, and predictions on validation or inference images. It covers installation, a training and validation workflow, and an inference-only workflow.

Get started

To use the integration, you must first install both ultralytics and wandb and confirm you’re using a supported version of ultralytics. Install ultralytics and wandb:

Command Line
Notebook

pip install --upgrade ultralytics==8.0.238 wandb

# or
# conda install ultralytics

!pip install --upgrade ultralytics==8.0.238 wandb

The development team tested the integration with ultralytics v8.0.238 and below. To report any issues with the integration, create a GitHub issue with the tag yolov8.

With both packages installed, you can move on to instrumenting an Ultralytics workflow with W&B.

Track experiments and visualize validation results

This section demonstrates a typical workflow that uses an Ultralytics model for training, fine-tuning, and validation, and that performs experiment tracking, model checkpointing, and visualization of the model’s performance using W&B. For more information about the integration, see Supercharging Ultralytics with W&B. To use the W&B integration with Ultralytics, import the wandb.integration.ultralytics.add_wandb_callback function. This callback is the entry point that registers W&B logging with the Ultralytics model.

import wandb
from wandb.integration.ultralytics import add_wandb_callback

from ultralytics import YOLO

Next, initialize the YOLO model of your choice, and invoke the add_wandb_callback function on it before performing inference with the model. Attaching the callback before training enables automatic logging during each epoch. This ensures that when you perform training, fine-tuning, validation, or inference, W&B automatically saves the experiment logs and the images, overlaid with both ground-truth and the respective prediction results using the interactive overlays for computer vision tasks, along with additional insights in a wandb.Table.

with wandb.init(project="ultralytics", job_type="train") as run:

    # Initialize YOLO Model
    model = YOLO("yolov8n.pt")

    # Add W&B callback for Ultralytics
    add_wandb_callback(model, enable_model_checkpointing=True)

    # Train/fine-tune your model
    # At the end of each epoch, predictions on validation batches are logged
    # to a W&B table with insightful and interactive overlays for
    # computer vision tasks
    model.train(project="ultralytics", data="coco128.yaml", epochs=5, imgsz=640)

With the callback attached and training started, your run now streams training metrics, model checkpoints, and per-epoch validation visualizations to your W&B project. Here’s how experiments tracked using W&B for an Ultralytics training or fine-tuning workflow look: Here’s how epoch-wise validation results are visualized using a W&B Table:

Visualize prediction results

This section demonstrates a typical workflow that uses an Ultralytics model for inference and visualizes the results using W&B. You can try out the code in Google Colab: Open in Colab. As with the training workflow, to use the W&B integration with Ultralytics, import the wandb.integration.ultralytics.add_wandb_callback function.

import wandb
from wandb.integration.ultralytics import add_wandb_callback

from ultralytics.engine.model import YOLO

Next, download a few images to test the integration on. You can use still images, videos, or camera sources. For more information about inference sources, see the Ultralytics docs.

!wget https://raw.githubusercontent.com/wandb/examples/ultralytics/colabs/ultralytics/assets/img1.png
!wget https://raw.githubusercontent.com/wandb/examples/ultralytics/colabs/ultralytics/assets/img2.png
!wget https://raw.githubusercontent.com/wandb/examples/ultralytics/colabs/ultralytics/assets/img4.png
!wget https://raw.githubusercontent.com/wandb/examples/ultralytics/colabs/ultralytics/assets/img5.png

With the test images in place, initialize a W&B run using wandb.init(). Next, initialize your desired YOLO model and invoke the add_wandb_callback function on it before you perform inference with the model. This ensures that when you perform inference, W&B automatically logs the images overlaid with your interactive overlays for computer vision tasks, along with additional insights in a wandb.Table.

# Initialize W&B Run
with wandb.init(project="ultralytics", job_type="inference") as run:
    # Initialize YOLO Model
    model = YOLO("yolov8n.pt")

    # Add W&B callback for Ultralytics
    add_wandb_callback(model, enable_model_checkpointing=True)

    # Perform prediction which automatically logs to a W&B Table
    # with interactive overlays for bounding boxes, segmentation masks
    model(
        [
            "img1.png",
            "img2.png",
            "img4.png",
            "img5.png",
        ]
    )

You don’t need to explicitly initialize a run using wandb.init() for a training or fine-tuning workflow. However, if the code involves only prediction, you must explicitly create a run.

After you run inference, W&B logs the predicted bounding boxes and segmentation masks to your W&B run as interactive overlays. Here’s how the interactive bounding box overlay looks: For more information, see the W&B image overlays guide.

​Get started

​Track experiments and visualize validation results

​Visualize prediction results

​More resources

Get started

Track experiments and visualize validation results

Visualize prediction results

More resources