kint8 量化模型封装代码

sunty

2022-03-21 d0a24896f95b4e060011852f80048ebfb0bf5f55

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
#ifndef NN_MATCHING_H 
#define NN_MATCHING_H 
 
#include "dataType.h" 
 
#include <map> 
 
//A tool to calculate distance; 
class NearNeighborDisMetric{ 
public: 
    enum METRIC_TYPE{euclidean=1, cosine}; 
    NearNeighborDisMetric(METRIC_TYPE metric, 
            float matching_threshold, 
            int budget); 
    ~NearNeighborDisMetric(); 
    DYNAMICM distance(const FEATURESS& features, const std::vector<uint64_t> &targets); 
    DYNAMICM cos_distance(const FEATURESS& features, std::map<uint64_t, FEATURESS> cos_samples); 
    //    void partial_fit(FEATURESS& features, std::vector<int> targets, std::vector<int> active_targets); 
    void partial_fit(std::vector<TRACKER_DATA>& tid_feats, std::vector<uint64_t>& active_targets); 
    float mating_threshold; 
 
private: 
    typedef Eigen::VectorXf (NearNeighborDisMetric::*PTRFUN)(const FEATURESS&, const FEATURESS&); 
    Eigen::VectorXf _nncosine_distance(const FEATURESS& x, const FEATURESS& y); 
    Eigen::VectorXf _nneuclidean_distance(const FEATURESS& x, const FEATURESS& y); 
 
    Eigen::MatrixXf _pdist(const FEATURESS& x, const FEATURESS& y); 
    Eigen::MatrixXf _cosine_distance(const FEATURESS & a, const FEATURESS& b, bool data_is_normalized = false); 
private: 
    PTRFUN _metric; 
    int budget; 
    std::map<uint64_t, FEATURESS > samples; 
}; 
 
#endif // NN_MATCHING_H