Trait FloatVisionOps 

pub trait FloatVisionOps: Backend {
    // Provided methods
    fn float_erode(
        input: FloatTensor<Self>,
        kernel: BoolTensor<Self>,
        opts: MorphOptions<Self, Float>,
    ) -> FloatTensor<Self> { ... }
    fn float_dilate(
        input: FloatTensor<Self>,
        kernel: BoolTensor<Self>,
        opts: MorphOptions<Self, Float>,
    ) -> FloatTensor<Self> { ... }
    fn nms(
        boxes: FloatTensor<Self>,
        scores: FloatTensor<Self>,
        options: NmsOptions,
    ) -> IntTensor<Self> { ... }
}

Vision ops on float tensors

Provided Methods

fn float_erode( input: FloatTensor<Self>, kernel: BoolTensor<Self>, opts: MorphOptions<Self, Float>, ) -> FloatTensor<Self>

Erodes an input tensor with the specified kernel.

fn float_dilate( input: FloatTensor<Self>, kernel: BoolTensor<Self>, opts: MorphOptions<Self, Float>, ) -> FloatTensor<Self>

Dilates an input tensor with the specified kernel.

fn nms( boxes: FloatTensor<Self>, scores: FloatTensor<Self>, options: NmsOptions, ) -> IntTensor<Self>

Perform Non-Maximum Suppression on bounding boxes.

Returns indices of kept boxes after suppressing overlapping detections. Boxes are processed in descending score order; a box suppresses all lower-scoring boxes with IoU > threshold.

Arguments

• boxes - Bounding boxes as [N, 4] tensor in (x1, y1, x2, y2) format
• scores - Confidence scores as [N] tensor
• options - NMS options (IoU threshold, score threshold, max boxes)

Returns

Indices of kept boxes as [M] tensor where M <= N

Dyn Compatibility

This trait is not dyn compatible.

In older versions of Rust, dyn compatibility was called "object safety".

Implementations on Foreign Types

impl FloatVisionOps for Flex

impl<B: FusionBackend + FloatVisionOps> FloatVisionOps for Fusion<B>

impl<R, F, I, BT> FloatVisionOps for CubeBackend<R, F, I, BT>

where R: CubeRuntime, F: FloatElement, I: IntElement, BT: BoolElement,

Implementors