Implementation:Ggml org Ggml Yolo detect

Summary

Primary API for YOLO object detection post-processing in the GGML framework. Converts raw neural network outputs into drawn bounding boxes with class labels on the input image.

Primary API

void detect(yolo_image & img, ggml_cgraph * gf, const yolo_model & model, float thresh, const vector<string> & labels, const vector<yolo_image> & alphabet)

Source: examples/yolo/yolov3-tiny.cpp:L455-479

Repository: https://github.com/ggml-org/ggml

Parameters

Parameter	Description
`img`	Input image for drawing
`gf`	Computed graph
`model`	YOLO model with anchors
`thresh`	Objectness threshold
`labels`	80 COCO class names
`alphabet`	Font images for labels

Related Functions

apply_yolo at L193-205: applies sigmoid activation, anchor decoding
get_yolo_detections at L235-265: extracts detections above threshold
do_nms_sort at L299-327: Non-Maximum Suppression
draw_detections at L340-386: draws bounding boxes and labels on image
save_image at examples/yolo/yolo-image.cpp:L49-64: saves result as JPEG

Outputs

JPEG image with drawn bounding boxes and class labels
Detection info printed to stdout

Configuration

NMS IoU threshold: 0.45
Classes: 80 COCO classes
Anchors: anchor sizes defined per scale

Dependencies

stb_image_write.h (for JPEG output)

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment