Skip to main content

detect_decode_forward

vision_rs::models::yolo::kernels::detect_decode

Function detect_decode_forward 

Source 
pub fn detect_decode_forward<T: Triton, D: Float, const BLOCK_A: i32>(
    boxes_ptr: T::Pointer<D>,
    anchor_x_ptr: T::Pointer<D>,
    anchor_y_ptr: T::Pointer<D>,
    strides_ptr: T::Pointer<D>,
    out_ptr: T::Pointer<D>,
    _B: i32,
    A: i32,
)
where
    T::I32Tensor: Tensor<i32, 1> + Comparison<i32, BoolTensor = T::BoolTensor>,
    T::Pointer<D>: AddOffsets<i32, 1, T::I32Tensor, Output = T::Tensor<T::Pointer<D>>>,
Expand description

Fused dist2bbox + stride-scale decode: LTRB distances → XYWH world coords.

Grid: B * cdiv(A, BLOCK_A) — one CTA per (batch element, anchor tile).