OBSERVE BEFORE YOU ACT

Find image

Search for a fixed image, trained object, pixel color or color coverage. Assign different behavior to found and not-found outcomes.

Find image editor · actual application, sample data
READ THE CONTROLS CLOSELY

Keep settings readable.

This is a 3× capture of the actual English module panel. Open it and select Inspect full size to read the details.

Find image · close-up · actual application, sample data

In an original game scene

Original demo world and actual Find image panel · original demo / illustrative composition

An illustrative composition with no third-party game assets. The product panel is real; the composition is not proof of an end-to-end commercial-game task.

When to use it

Search for a fixed image, trained object, pixel color or color coverage. Assign different behavior to found and not-found outcomes.

Settings and options

Fixed image

Crop a clean example from the screen or load a PNG. Extra background can make matching unreliable.

Trained object

Select a local object-library resource. For multiple detections, choose unique, highest confidence or nearest center.

Pixel color

Set a HEX color, tolerance and minimum pixel count. The same color may occur elsewhere in the interface.

Color coverage

Check a minimum/maximum percentage of the selected color in a region, for example a status bar.

Scope / region

Search the target window or screen. A smaller relevant region improves focus and reduces work.

Threshold and interval

Set confidence, attempts and polling interval. A similarity score is not a calibrated probability of correctness.

Wait until visible

Observe until the target appears. Normal attempt/timeout limits do not apply in this wait mode; Stop and Pause remain available.

Result actions

Choose Continue, Move, left/right/middle click, Stop or Go to group for the relevant outcome. Group actions need a destination.

Output variables

The Target prefix produces Target_Found, Target_X, Target_Y and Target_Score. Missing targets clear valid coordinates.

Test / supervisor

Preview the search before adding an action. A supervisor observes and triggers recovery without clicking during observation.

Hybrid tracking

With an appropriate model, YOLO11 ONNX detection can work with VOT local tracking. Without a model, a valid annotated example can support the exemplar tracker.

Observation versus action

A tracking box shows a located target. Clicking and steering are separately configured actions, not proof of a completed task.

Step-by-step setup

  1. Choose the search method and scope.
  2. Crop the image or select an object, then restrict the region.
  3. Configure both result paths on the result/test tab.
  4. Preview the correct target before saving.

Practical examples

EXAMPLE 01

Wait for a ready button

  1. Crop the stable Ready button.
  2. Enable Wait until visible and select its panel.
  3. Click when found, or Continue into a condition.
EXAMPLE 02

Monitor a status bar

  1. Choose Color coverage and isolate the bar.
  2. Set its color and percentage bounds.
  3. Branch on the result and test different fill levels.

Explore the interface

Find visual · all fields · actual application, sample data
Find image · Target · actual application, sample data
Find image · Result / trial · actual application, sample data
Visual search · TrainedObject · actual application, sample data
Visual search · PixelColor · actual application, sample data
Visual search · ColorCoverage · actual application, sample data

Tips and troubleshooting

  • For changing, rotating objects, consider the object library rather than one fixed image.
  • Before lowering the threshold, check crop, scale, DPI and region.
  • An indefinite wait cannot fix an incorrect region.

See shared step settings for targeting, results, timeouts and preconditions, and the formula library for expressions.

Next stepBuild a workflow with these modules ↗Resolve a problem ↗