Add configurable person box expansion

configs/sample_person_shoe_two_stage.json

@@ -45,7 +45,15 @@
           "num_classes": 80,
           "conf": 0.3,
           "nms": 0.45,
-          "class_filter": [0]
+          "class_filter": [0],
+          "bbox_expand": {
+            "enable": true,
+            "class_id": 0,
+            "left": 0.06,
+            "right": 0.06,
+            "top": 0.04,
+            "bottom": 0.16
+          }
         },
         {
           "id": "person_trk",

plugins/ai_yolo/ai_yolo_node.cpp

@@ -87,6 +87,15 @@ DetCoordContext BuildDetCoordContext(const Frame& frame, int model_input_w, int
     return ctx;
 }
 
+struct BboxExpandConfig {
+    bool enable = false;
+    int class_id = 0;
+    float left = 0.05f;
+    float right = 0.05f;
+    float top = 0.05f;
+    float bottom = 0.12f;
+};
+
 Rect DecodeToOutputRect(float x, float y, float w, float h, const DetCoordContext& ctx) {
     float ox = x;
     float oy = y;
@@ -115,6 +124,24 @@ Rect DecodeToOutputRect(float x, float y, float w, float h, const DetCoordContex
     return r;
 }
 
+Rect ExpandRect(const Rect& in, int img_w, int img_h, const BboxExpandConfig& cfg) {
+    if (!cfg.enable || img_w <= 0 || img_h <= 0) return in;
+
+    const float ex = in.w * cfg.left;
+    const float ey = in.h * cfg.top;
+    const float ew = in.w * (cfg.left + cfg.right);
+    const float eh = in.h * (cfg.top + cfg.bottom);
+
+    Rect out{};
+    out.x = std::max(0.0f, in.x - ex);
+    out.y = std::max(0.0f, in.y - ey);
+    out.w = std::min(static_cast<float>(img_w) - out.x, in.w + ew);
+    out.h = std::min(static_cast<float>(img_h) - out.y, in.h + eh);
+    out.w = std::max(0.0f, out.w);
+    out.h = std::max(0.0f, out.h);
+    return out;
+}
+
 inline int32_t ClipFloat(float val, float min_val, float max_val) {
     return static_cast<int32_t>(val <= min_val ? min_val : (val >= max_val ? max_val : val));
 }
@@ -677,6 +704,15 @@ public:
             }
         }
 
+        if (const SimpleJson* expand = config.Find("bbox_expand"); expand && expand->IsObject()) {
+            bbox_expand_.enable = expand->ValueOr<bool>("enable", false);
+            bbox_expand_.class_id = expand->ValueOr<int>("class_id", bbox_expand_.class_id);
+            bbox_expand_.left = expand->ValueOr<float>("left", bbox_expand_.left);
+            bbox_expand_.right = expand->ValueOr<float>("right", bbox_expand_.right);
+            bbox_expand_.top = expand->ValueOr<float>("top", bbox_expand_.top);
+            bbox_expand_.bottom = expand->ValueOr<float>("bottom", bbox_expand_.bottom);
+        }
+
         input_queue_ = ctx.input_queue;
         if (!input_queue_) {
             LogError("[ai_yolo] no input queue for node " + id_);
@@ -977,6 +1013,9 @@ private:
             det.cls_id = cls_id;
             det.score = obj_probs[i];
             det.bbox = DecodeToOutputRect(x1, y1, w, h, coord_ctx);
+            if (bbox_expand_.enable && det.cls_id == bbox_expand_.class_id) {
+                det.bbox = ExpandRect(det.bbox, coord_ctx.out_w, coord_ctx.out_h, bbox_expand_);
+            }
             det.track_id = -1;
 
             if (debug_det_ && det_result->items.size() < 5 && processed_ < 20) {
@@ -1127,6 +1166,9 @@ private:
             det.cls_id = cls_id;
             det.score = obj_probs[i];
             det.bbox = DecodeToOutputRect(x1, y1, w, h, coord_ctx);
+            if (bbox_expand_.enable && det.cls_id == bbox_expand_.class_id) {
+                det.bbox = ExpandRect(det.bbox, coord_ctx.out_w, coord_ctx.out_h, bbox_expand_);
+            }
             det.track_id = -1;
 
             if (debug_det_ && det_result->items.size() < 5 && processed_ < 20) {
@@ -1169,6 +1211,7 @@ private:
     bool stats_log_ = false;
     uint64_t stats_interval_ = 100;
     bool debug_det_ = false;
+    BboxExpandConfig bbox_expand_{};
 
     int64_t infer_interval_ms_ = 0;
     int64_t last_infer_pts_ms_ = 0;