update

README.md

@@ -71,7 +71,7 @@ OpenCV from JetPack repos (4.x) usually ships without CUDA dnn. To use CUDA, bui
 - `RETURN_BOXES` (default `true`)
 - `USE_CUDA_DNN` (default `false`, set `true` to attempt CUDA backend for EAST if OpenCV built with CUDA)
 - `DEBUG_VISUAL` (default `false`; if true, response includes `debug_image_base64`. Per-request: add `?debug=1` or form field `debug=1`.)
-- `SHOW_GUI` (default `false`; if true and OpenCV GUI available, service opens a window with overlays. Requires X/Wayland/VDI that supports GUI.)
+- `SHOW_GUI` (default `false`; if true and OpenCV GUI available, service opens a text-centric window: left = detected text regions, right = OCR block text/conf. Requires X/Wayland/VDI that supports GUI.)
 
 ## Run
 ```bash

app.py

@@ -235,6 +235,90 @@ def build_debug_image(img_rgb, boxes, blocks, objects, target_width=1024, return
     except Exception:
         return None, None
 
+
+def build_text_gui_image(img_rgb, boxes, blocks, target_width=1600):
+    """Build a text-centric GUI frame: regions on left, OCR lines on right."""
+    if cv2 is None or img_rgb is None:
+        return None
+    try:
+        src = cv2.cvtColor(img_rgb, cv2.COLOR_RGB2BGR)
+        h, w = src.shape[:2]
+        frame_w = target_width if target_width and target_width > 0 else 1600
+        if frame_w < 900:
+            frame_w = 900
+
+        left_w = int(frame_w * 0.64)
+        scale = float(left_w) / float(max(1, w))
+        left_h = max(360, int(h * scale))
+        left = cv2.resize(src, (left_w, left_h), interpolation=cv2.INTER_AREA if scale < 1.0 else cv2.INTER_CUBIC)
+
+        # text detector regions
+        for idx, b in enumerate(boxes or []):
+            try:
+                x1, y1, x2, y2 = [int(v * scale) for v in b]
+            except Exception:
+                continue
+            cv2.rectangle(left, (x1, y1), (x2, y2), (255, 90, 0), 2)
+            cv2.putText(left, "T{}".format(idx + 1), (x1, max(12, y1 - 4)),
+                        cv2.FONT_HERSHEY_SIMPLEX, 0.45, (255, 90, 0), 1, cv2.LINE_AA)
+
+        # OCR blocks (where text was actually extracted)
+        for idx, blk in enumerate(blocks or []):
+            try:
+                x1, y1, x2, y2 = [int(v * scale) for v in blk.get("bbox", [0, 0, 0, 0])]
+            except Exception:
+                continue
+            cv2.rectangle(left, (x1, y1), (x2, y2), (0, 160, 255), 2)
+            cv2.putText(left, "R{}".format(idx + 1), (x1, min(left_h - 8, y2 + 14)),
+                        cv2.FONT_HERSHEY_SIMPLEX, 0.45, (0, 160, 255), 1, cv2.LINE_AA)
+
+        right_w = frame_w - left_w
+        if right_w < 320:
+            right_w = 320
+            left = cv2.resize(left, (frame_w - right_w, left_h))
+            left_w = left.shape[1]
+        right = np.ones((left_h, right_w, 3), dtype=np.uint8) * 245
+
+        y = 28
+        line_h = 22
+        cv2.putText(right, "OCR text panel", (14, y), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (30, 30, 30), 2, cv2.LINE_AA)
+        y += line_h + 4
+        cv2.putText(right, "detected regions: {}".format(len(boxes or [])), (14, y),
+                    cv2.FONT_HERSHEY_SIMPLEX, 0.55, (70, 70, 70), 1, cv2.LINE_AA)
+        y += line_h
+        cv2.putText(right, "recognized blocks: {}".format(len(blocks or [])), (14, y),
+                    cv2.FONT_HERSHEY_SIMPLEX, 0.55, (70, 70, 70), 1, cv2.LINE_AA)
+        y += line_h + 4
+        cv2.line(right, (12, y), (right_w - 12, y), (200, 200, 200), 1)
+        y += 18
+
+        if not blocks:
+            cv2.putText(right, "No OCR text blocks", (14, y),
+                        cv2.FONT_HERSHEY_SIMPLEX, 0.56, (100, 100, 100), 1, cv2.LINE_AA)
+        else:
+            visible = blocks[:10]
+            max_chars = max(18, int((right_w - 24) / 8))
+            for idx, blk in enumerate(visible):
+                conf = blk.get("conf", -1)
+                conf_str = "{:.1f}".format(conf) if isinstance(conf, (float, int)) and conf >= 0 else "n/a"
+                header = "#{:02d} conf={}".format(idx + 1, conf_str)
+                cv2.putText(right, header, (14, y), cv2.FONT_HERSHEY_SIMPLEX, 0.50, (20, 20, 20), 1, cv2.LINE_AA)
+                y += line_h - 2
+                raw = (blk.get("text", "") or "").replace("\n", " ").replace("\r", " ").strip()
+                txt = " ".join(raw.split())
+                if len(txt) > max_chars:
+                    txt = txt[:max_chars - 1] + "..."
+                cv2.putText(right, txt if txt else "<empty>", (14, y),
+                            cv2.FONT_HERSHEY_SIMPLEX, 0.52, (30, 30, 30), 1, cv2.LINE_AA)
+                y += line_h
+                if y > left_h - 18:
+                    break
+
+        frame = np.concatenate([left, right], axis=1)
+        return frame
+    except Exception:
+        return None
+
 # ---- Routes ----
 
 @app.route('/health', methods=['GET'])
@@ -382,9 +466,18 @@ def image2text():
             target_width=target_w,
             return_image=SHOW_GUI,
         )
-        if SHOW_GUI and cv2 is not None and debug_image_arr is not None:
+        if SHOW_GUI and cv2 is not None:
             try:
-                _gui_show(debug_image_arr)
+                gui_frame = build_text_gui_image(
+                    img_rgb,
+                    boxes,
+                    blocks if RETURN_BOXES else [],
+                    target_width=GUI_TARGET_WIDTH,
+                )
+                if gui_frame is None:
+                    gui_frame = debug_image_arr
+                if gui_frame is not None:
+                    _gui_show(gui_frame)
             except Exception:
                 pass