Identify objects in a photo, draw their bounding boxes, and count them by category.
The Florence-2 detection task returns categories and coordinates. UtiliTools Pro draws these regions on the photograph, counts objects by label and allows you to download the annotated result.
Locate multiple items in a photo.
Draw each region detected on the image.
Group labels and see how many objects were detected.
Download the photo with the boxes and labels.
The photo is analyzed on the device.
Run the model locally via WebAssembly.
Select a picture.
Click “Detect objects”.
Review the bounding boxes and the count for each category.
Download the annotated image in PNG.
This feature uses Florence-2 Base FT in ONNX format. The model handles OCR, image descriptions, and detection through different prompts, so the browser can reuse stored model files after the first download.
Florence-2 runs through WebAssembly for broader browser compatibility. The initial download is much larger than for a typical web tool, but cached files can be reused.
Florence-2 recognizes many common categories, but there is no fixed, guaranteed list, and it can make mistakes in complex scenes.
This Florence-2 task returns labels and bounding boxes, but no confidence percentage that we can reliably display.
The model labels are usually in English.
No. It is processed locally after downloading the model files.
Yes. The tool generates a PNG annotated with visible detections.
