DetectionTypes.h File

Detection result types and helpers for decoding model outputs. More...

Included Headers

#include "pipeline/TensorCore.h" #include <cstdint> #include <string> #include <vector>

Namespaces Index

namespace	simaai
namespace	neat

Classes Index

struct	Box
	One axis-aligned detection produced by BoxDecode. More...
struct	BoxDecodeResult
	Parsed BoxDecode output paired with its raw byte buffer. More...

Description

Detection result types and helpers for decoding model outputs.

Defines the structured output types produced after the BoxDecode postprocess stage runs. Box is a single detection (axis-aligned bbox + class + score); BoxDecodeResult bundles the parsed list with the raw byte payload it came from. The free functions parse byte-encoded BBOX tensors back into typed Box records for application code.

See Also

BoxDecodeType for the decode-family selection.

See Also

StageRun.h for the BoxDecode stage entry point.

File Listing

The file content with the documentation metadata removed is:

1

15#pragma once

16

17#include "pipeline/TensorCore.h"

18

19#include <cstdint>

20#include <string>

21#include <vector>

22

23namespace simaai::neat {

24

35struct Box {

36 float x1 = 0.0f;

37 float y1 = 0.0f;

38 float x2 = 0.0f;

39 float y2 = 0.0f;

40 float score = 0.0f;

41 int class_id = -1;

42};

43

52struct BoxDecodeResult {

53 std::vector<Box> boxes;

54 std::vector<uint8_t> raw;

55};

56

67std::vector<Box> parse_bbox_bytes(const std::vector<uint8_t>& bytes, int img_w, int img_h,

68 int expected_topk, bool strict);

69

83BoxDecodeResult decode_bbox_tensor(const simaai::neat::Tensor& tensor, int img_w, int img_h,

84 int expected_topk, bool strict);

85

86} // namespace simaai::neat

---

Generated via doxygen2docusaurus 2.0.0 by Doxygen 1.9.1.