Workflow Blocks

Predict the location of objects with bounding boxes.

Predict the shape, size, and location of objects.

Generate an embedding of an image or string.

Apply a single tag to an image.

Apply multiple tags to an image.

Predict skeletons on objects.

Run Anthropic Claude model with vision capabilities.

Run Google's Gemini model with vision capabilities.

Run OpenAI's GPT-4 with vision capabilities.

Run Florence-2 on an image

Run a zero-shot object detection model.

Run a self-hosted vision language model.

Convert bounding boxes to polygons, or run SAM2 on an entire image to generate a mask.

Compare CLIP image and text embeddings.

Extract text from an image using optical character recognition.

Detect and read barcodes in an image.

Detect and read QR codes in an image.

Detect faces and estimate gaze direction

Use segmentation masks to inpaint objects within an image.

Detect text in images using Google Vision API

Run a large multimodal model such as ChatGPT-4v or CogVLM.

Run a large multimodal model such as ChatGPT-4v or CogVLM for classification.

generate new images from text, or create variations of existing images.

Run Llama model with Vision capabilities

Draw a box around detected objects in an image.

Draw a polygon around detected objects in an image.

Draw labels on an image at specific coordinates based on provided detections.

Visualize both single-label and multi-label classification predictions with customizable display options.

Apply a mask to cover all areas outside the detected regions in an image.

Blur detected objects in an image.

Draw a circle around detected objects in an image.

Paint a solid color on detected objects in an image.

Draw the corners of detected objects in an image.

Draw scaled up crops of detections on the scene.

Draw ellipses that highlight detected objects in an image.

Paint a halo around detected objects in an image.

Apply a mask over detected objects in an image.

Pixelate detected objects in an image.

Draw triangle markers on an image at specific coordinates based on provided detections.

Apply a mask over a line zone in an image.

Apply a mask over a polygon zone in an image.

Visualize the difference between two models' detections.

Draw traces based on detections tracking results.

Draw a reference path in the image.

Draw keypoints on detected objects in an image.

Shows an array of images in a grid.

Conditionally stop execution of a branch.

Conditionally filter out model predictions.

Limits the rate at which a branch of the Workflow will run.

Allow the execution of workflow to proceed if the input value has changed.

Combine predictions from multiple detections models to make a decision about object presence.

Save images and predictions to your Roboflow Dataset.

Send a request to a remote API with Workflow results.

Create CSV files with specified columns.

Save data to a local file.

Aggregate workflow data to produce time-based statistics.

Send notification via e-mail.

Send notification via Slack.

Send notification via Twilio SMS service.

Crop an image using bounding boxes from a detection model.

Crop an image using fixed pixel coordinates.

Crop an image proportional (%) to its dimensions.

Apply a padding around the width and height of detections.

Apply transformations on detected bounding boxes.

Find the minimal bounding box surrounding the detected polygon.

Stitch two images by common parts.

Resize, flip, or rotate an image.

Locate instances of a given template within a specified image.

Get the dominant color of an image in RGB format.

Draw dots on an image at specific coordinates based on provided detections.

Count the number of pixels that match a specific color within a given tolerance.

Compare SIFT descriptors from multiple images.

Find and count the contours on an image.

Apply SIFT to an image.

Apply a blur to an image.

Apply a threshold to an image.

Convert an RGB image to grayscale.

Calculate a score to indicate how well-focused a camera is.

Measure the dimensions of objects in relation to a reference object.

Calculate the distance between two bounding boxes on a 2D plane.

Track and update object positions across video frames using ByteTrack.

Track object time in zone.

Count detections passing a line.

Calculate Fréchet distance of object from the reference path.

Apply a smoothing algorithm to reduce noise and flickering across video frames.

Returns an array of the last `length` values passed to it.

Identify outlier embeddings compared to prior data.

Identify changes compared to prior data via embeddings.

Define a variable from model predictions, such as the class names, confidences, or number of detections.

Create a specific output based on defined input variables and configured rules.

Adjust detection coordinates from a polygon-defined plane to a straight rectangular plane with specified width and height.

Combines OCR detection results into a coherent text string by organizing detections spatially.

Simplify polygons so they are geometrically convex and contain only the requested amount of vertices.

Calculate the cosine similarity between two embeddings.

Replace classes of detections with classes predicted by a chained classification model.

Parse raw string into JSON.

Parse a raw string into a classification prediction.

Parses raw string into object-detection prediction.

Collapse dimensionality by aggregating nested data into a single list.

Take the first non-empty data element or the configured default value.

Add custom metadata to the Roboflow Model Monitoring dashboard.

Periodically report an aggregated sample of inference results to Roboflow Model Monitoring.

Tile the input image into a list of smaller images to perform small object detection.

Merges detections made against multiple pieces of input image into single detection.

Fetches a previously stored value from a cache entry.

Stores a value in a cache entry for later retrieval.

Fetch secrets from environmental variables.