client_stream_message.proto

syntax = "proto3";

// BBox2DType represents a single bounding box with semantic information
message BBox2DType {
    uint32 semanticId = 1;      // Unique identifier for the semantic class
    int32 xMin = 2;             // Left coordinate of the bounding box
    int32 yMin = 3;             // Top coordinate of the bounding box
    int32 xMax = 4;             // Right coordinate of the bounding box
    int32 yMax = 5;             // Bottom coordinate of the bounding box
    float occlusionRatio = 6;   // Ratio indicating how much of the object is occluded (0-1)
}

// BBox2DInfo contains metadata for a collection of bounding boxes
message BBox2DInfo {
    map<string, string> idToLabels = 1;  // Maps semantic IDs to human-readable labels
    repeated int32 bboxIds = 2;          // List of bounding box identifiers
}

// BBox2D represents a complete set of bounding boxes with their metadata
message BBox2D {
    repeated BBox2DType data = 1;  // Collection of bounding boxes
    BBox2DInfo info = 2;           // Metadata for the bounding boxes
}

// Clock contains timing information for synchronization
message Clock {
    double sim_dt = 1;     // Simulation delta time (seconds)
    double sys_dt = 2;     // System delta time (seconds)
    double sim_time = 3;   // Current simulation time (seconds)
    double sys_time = 4;   // Current system time (seconds)
}

// Camera contains camera parameters and transformation matrices
message Camera {
    repeated double view_matrix_ros = 1;    // Flattened 4x4 matrix (size 16) for camera pose in ROS format
    repeated double camera_scale = 2;       // Size 3 vector for camera scaling factors
    repeated double intrinsics_matrix = 3;  // Flattened 3x3 matrix (size 9) for camera intrinsics
}

// ClientStreamMessage is the main message containing all data transmitted
message ClientStreamMessage {
    BBox2D bbox2d = 1;        // Bounding box data
    Clock clock = 2;          // Timing information
    Camera camera = 3;        // Camera parameters
    bytes color_image = 4;    // RGB image data (encoded)
    bytes depth_image = 5;    // Depth image data (encoded)
}