Results¶

All models return results in a unified Y structure.

Quick Start

let ys: Vec<Y> = model.run(&images)?;

Y Structure¶

Field	Type	Description
`hbbs`	`Vec<Hbb>`	Horizontal bounding boxes
`obbs`	`Vec<Obb>`	Oriented bounding boxes
`masks`	`Vec<Mask>`	Segmentation masks
`polygons`	`Vec<Polygon>`	Contours
`keypoints`	`Vec<Keypoint>`	Keypoints
`keypointss`	`Vec<Vec<Keypoint>>`	Multiple keypoint sets
`probs`	`Vec<Prob>`	Classification probabilities
`texts`	`Vec<Text>`	OCR/VLM text
`embedding`	`X`	Feature embeddings
`extra`	`HashMap<String, X>`	Model-specific data
`images`	`Vec<Image>`	Output images

Common Methods¶

All result types have metadata methods via impl_meta_methods!:

Method	Returns	Description
`id()`	`Option<usize>`	Class ID
`name()`	`Option<&str>`	Class name
`confidence()`	`Option<f32>`	Confidence score
`track_id()`	`Option<usize>`	Tracking ID
`uid()`	`usize`	Unique instance ID

Hbb (Horizontal Bounding Box)¶

Geometry Methods¶

Method	Returns	Description
`xmin()` / `ymin()`	`f32`	Top-left corner
`xmax()` / `ymax()`	`f32`	Bottom-right corner
`cx()` / `cy()`	`f32`	Center coordinates
`width()` / `height()`	`f32`	Box dimensions
`xyxy()`	`(f32, f32, f32, f32)`	[x1, y1, x2, y2]
`xywh()`	`(f32, f32, f32, f32)`	[x, y, w, h]
`cxywh()`	`(f32, f32, f32, f32)`	[cx, cy, w, h]
`cxcyah()`	`(f32, f32, f32, f32)`	[cx, cy, ar, h]
`area()`	`f32`	Box area
`perimeter()`	`f32`	Box perimeter
`is_square()`	`bool`	Check if square

Operations¶

Method	Description
`intersect(&other)`	Intersection area
`union(&other)`	Union area
`iou(&other)`	IoU ratio
`contains(&other)`	Contains check
`to_polygon()`	Convert to Polygon

Constructors¶

Method	Description
`from_xyxy(x1, y1, x2, y2)`	From corners
`from_xywh(x, y, w, h)`	From position + size
`from_cxcywh(cx, cy, w, h)`	From center + size

Example

for hbb in y.hbbs() {
    let [x1, y1, x2, y2] = [hbb.xmin(), hbb.ymin(), hbb.xmax(), hbb.ymax()];
    println!("{}: {:.2}%", hbb.name().unwrap_or("?"), hbb.confidence().unwrap_or(0.0) * 100.0);
}

Obb (Oriented Bounding Box)¶

Geometry Methods¶

Method	Returns	Description
`coords()`	`&[[f32; 2]; 4]`	4 vertices (CCW)
`area()`	`f32`	Polygon area
`top()` / `left()`	`[f32; 2]`	Extreme points
`bottom()` / `right()`	`[f32; 2]`	Extreme points
`is_hbb()`	`bool`	Check axis-aligned

Operations¶

Method	Description
`intersect(&other)`	Intersection area (Sutherland-Hodgman)
`union(&other)`	Union area
`iou(&other)`	IoU ratio
`to_polygon()`	Convert to Polygon

Constructors¶

Method	Description
`from_cxcywhd(cx, cy, w, h, degrees)`	Center + size + angle (deg)
`from_cxcywhr(cx, cy, w, h, radians)`	Center + size + angle (rad)

Mask¶

Method	Returns	Description
`width()` / `height()`	`u32`	Dimensions
`dimensions()`	`(u32, u32)`	(w, h)
`to_vec()`	`Vec<u8>`	Raw data
`polygon()`	`Option<Polygon>`	Largest contour
`polygons()`	`Vec<Polygon>`	All contours

Example

for mask in y.masks() {
    println!("Mask: {}x{}", mask.width(), mask.height());
    if let Some(poly) = mask.polygon() {
        println!("Area: {}", poly.area());
    }
}

Polygon¶

Method	Returns	Description
`count()`	`usize`	Number of points
`points()`	`Vec<[f32; 2]>`	Clone of coords
`exterior()`	`&[[f32; 2]]`	Coords slice
`is_closed()`	`bool`	Check if closed
`area()`	`f64`	Shoelace formula
`perimeter()`	`f64`	Euclidean length
`centroid()`	`Option<(f32, f32)>`	Center point

Operations¶

Method	Description
`intersect(&other)`	Intersection area
`union(&other)`	Union area
`hbb()`	Bounding box
`obb()`	Minimum rotated rect
`convex_hull()`	Convex hull
`simplify(eps)`	RDP simplification
`resample(n)`	Add points on edges
`unclip(delta, w, h)`	Expand polygon

Keypoint¶

Method	Returns	Description
`x()` / `y()`	`f32`	Coordinates
`xy()`	`(f32, f32)`	(x, y) tuple
`is_origin()`	`bool`	Check (0, 0)
`distance_from(&other)`	`f32`	Euclidean distance
`distance_from_origin()`	`f32`	Distance from (0,0)
`rotate(rad)`	`Self`	Rotate around origin
`cross(&other)`	`f32`	Cross product

Operators¶

+, -, *, / with f32 or Keypoint

Prob (Classification)¶

Method	Description
`new_probs(probs, names, k)`	Create top-k probs

Text (OCR/VLM)¶

Method	Returns	Description
`text()`	`&str`	Text content

Access Patterns¶

Borrow (Reference)¶

Example

let ys = model.run(&xs)?;

for y in &ys {
    for hbb in y.hbbs() {
        println!("Box: {:?}, Conf: {:?}", hbb.xyxy(), hbb.confidence());
    }
}

Consume (Ownership)¶

Example

let ys = model.run(&xs)?;

for y in ys {
    let boxes: Vec<Hbb> = y.hbbs;
}