Ultralytics Snippets for Visual Studio Code

A Python snippets extension for VS Code to assist with development using the Ultralytics package. These snippets will help you code with Ultralytics faster and help provide some boilerplate examples to test out. Open an Issue or a Pull Request to have your snippet added! 🚀 Also works with neovim!

NOTE: Tab-completion used to complete snippet and quickly hop into fields.

Updates

• 2024-10-03 :: Ultralytics YOLO11 added! YOLO11 will now be the default model.
• 2024-08-26 :: Ultralytics VS Code Snippets integration documentation page.
• 2024-08-14 :: Extension size is now ~100x smaller.
• 2024-08-03 :: Tracking example and KWARGs snippets added.
• 2024-07-30 :: SAM2 examples included.
• 2024-07-24 :: Lots of new snippets and updates.
• 2024-07-17 :: Adds YOLO-world snippet with custom prompts, updates reference links, and adds keyword argument.
• 2024-07-12 :: Model export snippet added and adds keyword argument.

Installation

From the Web

Method-1

1. Visit the VS Code Extension Marketplace by going to https://marketplace.visualstudio.com/VSCode and search for Ultralytics-Snippets.

Method-2

1. Follow this link to visit the extension page directly.

2. Click the "Install" button and allow your browser to launch a VS Code session.

   

From VS Code

1. Navigate to the Extensions menu, and search for Ultralytics-Snippets.

2. Click the "Install" button.

   

From the CLI

You can also install the latest version of the Ultralytics-Snippets VS Code extension using the following command.

code --install-extension ultralytics.ultralytics-snippets

Snippets Syntax

All snippets use the format:

{PREFIX}.{ROOT}*-{DESCRIPTOR}

• {PREFIX} is always ultra

• {ROOT} denotes a common "root" verb or noun such as import or results. There will always be at least one root common with other snippets, but it's possible there could be more than one, such as result-boxes.

• {DESCRIPTOR} will be related to the snippet functionality and all words will be separated with hyphens - the snippet alias.

Import

Import snippets are for common objects that would be imported from the Ultralytics library.

Alias	Description
ultra.import-model	Add line to import Ultralytics YOLO
ultra.import-assets	Import Ultralytics ASSETS directory constant.
ultra.import-results	Import Ultralytics Results class (usually for type hinting).
ultra.import-annotator	Import Ultralytics auto_annotate function.
ultra.import-coco2yolo	Import Ultralytics function to convert annotations from COCO to YOLO format.
ultra.import-bbox2seg	Import Ultralytics function to convert horizontal bounding boxes to segmentation contours.
ultra.import-seg2bbox	Import Ultralytics function to convert segmentation contours into horizontal bounding boxes.
ultra.import-box-convert	Import Ultralytics function for converting bounding box coordinates.
ultra.import-formats	Import Ultralytics supported file formats constant.
ultra.import-task-result	Import task-based results class (generally helpful for type hinting).

Snippet Example

ultra.import-model Snippet

Drop-down select available for Model class to import.

Model = "YOLO"
from ultraltyics import f"{Model}"  # not intended to represent valid code

Results

These snippets will provide shortcuts for working with ultralytics.engine.results.Results objects returned from model inference. See the Working with Results of the documentation and the Results class reference page for more information.

Alias	Description
ultra.result-class-str	Convert class indices to class string names for a single image result.
ultra.result-data	Get result data array of detections from a single image result.
ultra.result-loop	Iterate prediction results from an Ultralytics model.
ultra.result-box-xyxy	Get pixel-value (x1, y1, x2, y2) bounding box coordinates from a single image result.
ultra.result-box-xywh	Get pixel-value (x-center, y-center, w, h) bounding box coordinates from a single image result.
ultra.result-mask-binary	Get binary segmentation masks from a single image result. NOTE: [N, H, W] shape, with inference H, W dimensions.
ultra.result-mask-contours	Get segmentation contours with pixel value xy or normalized xyn coordinates.
ultra.result-obb-xywhr	Get OBB rotated bounding boxes in pixel value [x, y, w, h, r] coordinates as torch.Tensor array.
ultra.result-orig-image	Get original image from a single image result.
ultra.result-filter-class	Filter prediction results by class ID. Using classes keyword argument for prediction should be preferred.

Snippet Example

ultra.result-loop Snippet

for result in results:
    result.boxes.data  # torch.Tensor array

NOTE: results is a placeholder and can be modified to match existing naming schema.

Models

Shortcuts for initializing pretrained Ultralytics models, like YOLOv8.

Alias	Description	Reference
ultra.yolo-model	Shortcut to initialize YOLO model.	YOLOv5, YOLOv8, YOLOv9, YOLOv10, YOLO-World
ultra.yolo-export	Shortcut to export YOLO model weights.	Model Export
ultra.sam-model	Shortcut to initialize SAM.	SAM
ultra.mobileam-model	Shortcut to initialize MobileSAM.	Mobile SAM
ultra.fastam-model	Shortcut to initialize FastSAM.	FastSAM
ultra.nas-model	Shortcut to initialize YOLO-NAS model.	YOLO-NAS
ultra.rtdetr-model	Shortcut to initialize RTDETR model.	RTDETR
ultra.yolo-world-model	Shortcut to initialize YOLO-world model, with class prompts.	YOLO-World
ultra.sam2-bboxes	Shortcut to initialize YOLO-World model with text prompts.	SAM2
ultra.sam2-points	Shortcut to initialize YOLO-World model with text prompts.	SAM2

Snippet Example

ultra.yolo-model Snippet

Drop-down select available for version, scale, and task, equivalent Python code shown below

version = 8
scale = "s"
task = "."  # detect
model = YOLO(f"yolov{version}{scale}{task}pt")

version = 9
scale = "e"
task = "-seg."  # segment
model = YOLO(f"yolov{version}{scale}{task}pt")

NOTE: It will be possible to create combinations that aren't available, such as yolov8n-worldv2.pt. User is responsible for creating a valid configuration per documentation.

Utilities

Alias	Description	Reference
ultra.util-auto-annotate	Use Ultralytics auto_annotate function to generate annotations.	auto_annotator fucntion
ultra.util-annotator	Use Ultralytics Annotator class to draw box annotations	Annotator class
ultra.util-make-divisible	Use Ultralytics make_divisible function to make a number divisible by another.	make_divisible function
ultra.util-callback	Shortcut for adding custom model callback for a defined function.	callbacks

Snippet Example

ultra.auto-annotate Snippet

from ultralytics.data.annotator import auto_annotate

auto_annotate(data="", det_model="yolov8n.pt", sam_model="sam_b.pt", device="cuda", output_dir="")

NOTE: Each function argument will be a "field" that can be tabbed into and changed. The det_model, sam_model, and device arguments will have options for default models, but can be cleared to input custom strings instead.

Examples

The Example snippets are more "complete" blocks of code that can be used for boilerplate demonstrations.

Prefix	Description
ultra.example-predict-filter-class	Ultralytics basic YOLO object detection predict with filtered classes example.
ultra.example-result-filter-class	Filter prediction results by class ID. Using "classes" keyword argument for prediction should be preferred.
ultra.example-yolo-predict	Setup Ultralytics YOLO to perform predict (simple).
ultra.example-yolo-val	Setup Ultralytics YOLO to perform validation (simple).
ultra.example-yolo-train	Setup Ultralytics YOLO to perform training (simple).
ultra.example-yolo-predict-kwords	Setup Ultralytics YOLO to perform inference, show all inference keyword arguments and their default values.
ultra.example-yolo-train-kwords	Setup Ultralytics YOLO for training, with all keyword arguments and their default values.
ultra.example-sam-predict	Setup Ultralytics SAM to perform inference (simple).
ultra.example-sam2	Example showing use of SAM2 with bounding box and point prompts.
ultra.example-mobile-sam-predict	Setup Ultralytics MobileSAM to perform inference (simple).
ultra.example-fast-sam-predict	Setup Ultralytics FastSAM to perform inference (simple).
ultra.example-nas-predict	Setup Ultralytics NAS to perform inference (simple).
ultra.example-rtdetr-predict	Setup Ultralytics RT-DETR to perform inference (simple).
ultra.example-callback	Example showing how to add a custom callback function.
ultra.example-track-loop-persist	Example of how to open video, loop frames, and maintain tracked object IDs.
ultra.example-track-kwords	Example showing all keyword arguments available for track mode.

Snippet Example

ultra.example-predict Snippet

from ultralytics import YOLO, ASSETS

model = YOLO("yolov8n.pt", task="detect")
results = model(source=ASSETS / "bus.jpg")

for result in results:
    print(result.boxes.data)
    # result.show()  # uncomment to view each result image

NOTE: Here, the only configurable option is the model scale which can be any one of: n, s, m, l, or x.

KWARGS

Use these to insert the various model methods defined in modes with all keyword arguments, default values, and commented descriptions quickly into your code. Includes model as default variable, but is an editable field accessible using tab stops.

Prefix	Description	Reference
ultra.kwargs-predict	Snippet using model predict method, including all keyword arguments and defaults.	predict
ultra.kwargs-train	Snippet using model train method, including all keyword arguments and defaults.	train
ultra.kwargs-track	Snippet using model track method, including all keyword arguments and defaults.	track
ultra.kwargs-val	Snippet using model val method, including all keyword arguments and defaults.	val

Snippet Example

ultra.kwargs-predict

model.predict(
    source=src,           # (str, optional) source directory for images or videos
    imgsz=640,            # (int | list) input images size as int or list[w,h] for predict
    conf=0.25,            # (float) minimum confidence threshold
    iou=0.7,              # (float) intersection over union (IoU) threshold for NMS
    vid_stride=1,         # (int) video frame-rate stride
    stream_buffer=False,  # (bool) buffer all streaming frames (True) or return the most recent frame (False)
    visualize=False,     # (bool) visualize model features
    augment=False,       # (bool) apply image augmentation to prediction sources
    agnostic_nms=False,  # (bool) class-agnostic NMS
    classes=None,        # (int | list[int], optional) filter results by class, i.e. classes=0, or classes=[0,2,3]
    retina_masks=False,  # (bool) use high-resolution segmentation masks
    embed=None,          # (list[int], optional) return feature vectors/embeddings from given layers
    show=False,          # (bool) show predicted images and videos if environment allows
    save=True,           # (bool) save prediction results
    save_frames=False,   # (bool) save predicted individual video frames
    save_txt=False,      # (bool) save results as .txt file
    save_conf=False,     # (bool) save results with confidence scores
    save_crop=False,     # (bool) save cropped images with results
    stream=False,        # (bool) for processing long videos or numerous images with reduced memory usage by returning a generator
    verbose=True,        # (bool) enable/disable verbose inference logging in the terminal
)

Use with neovim

It's possible to use VS Code snippets by installing the LuaSnip repo and then adding the following line into your configuration of LuaSnip:

require("luasnip.loaders.from_vscode").lazy_load({ paths = { "./ultralytics-snippets/" }, include = { "python" } })

Make sure that the path "./ultralytics-snippets/" is valid for your install location.

Note

If the snippets don't work, try removing the comment lines at the top of each JSON file. These are ignored by VS Code, but might not be ignored by neovim or LuaSnip.