zone_alarm/src/rga_processor.h

#pragma once
#include <cstdint>
#include <vector>
#include "rtsp_capture.h"

class RgaProcessor {
public:
    RgaProcessor() = default;
    ~RgaProcessor();
    
    // Initialize with target input size for the model (e.g. 640x640)
    bool init(int target_w, int target_h);
    
    // Convert NV12 to RGB and resize to target size
    // Returns RGB data in NHWC format, normalized to [0,1] as float
    // Also returns the scale factor and padding offsets for coordinate mapping
    bool process(const Nv12Frame& nv12, 
                 std::vector<float>& output,
                 float& scale, int& pad_x, int& pad_y);
    
    // Get the scale factor used in last process() call
    float lastScale() const { return last_scale_; }
    int lastPadX() const { return last_pad_x_; }
    int lastPadY() const { return last_pad_y_; }
    
    // Direct NV12->RGB without normalization (for display/JPEG encoding)
    bool processToRgb(const Nv12Frame& nv12, std::vector<uint8_t>& rgb_out,
                      int& out_w, int& out_h);

private:
    int target_w_ = 640;
    int target_h_ = 640;
    float last_scale_ = 1.0f;
    int last_pad_x_ = 0;
    int last_pad_y_ = 0;
    
    // Software fallback methods
    void softwareNv12ToRgb(const Nv12Frame& nv12, std::vector<uint8_t>& rgb,
                           int out_w, int out_h);
};