Use dora metadata to reduce complexity as well as enable zero-copy image transfering

node-hub/opencv-plot/opencv_plot/main.py

@@ -1,6 +1,6 @@
 import os
 import argparse
-from cv2 import cv2  # Importing cv2 this way remove error warnings``
+import cv2
 
 import numpy as np
 import pyarrow as pa
@@ -129,25 +129,24 @@ def main():
             event_id = event["id"]
 
             if event_id == "image":
-                arrow_image = event["value"][0]
+                storage = event["value"]
+
+                metadata = event["metadata"]
+                encoding = metadata["encoding"]
+                width = metadata["width"]
+                height = metadata["height"]
 
-                encoding = arrow_image["encoding"].as_py()
                 if encoding == "bgr8":
                     channels = 3
                     storage_type = np.uint8
                 else:
                     raise RuntimeError(f"Unsupported image encoding: {encoding}")
 
-                image = {
-                    "width": np.uint32(arrow_image["width"].as_py()),
-                    "height": np.uint32(arrow_image["height"].as_py()),
-                    "encoding": encoding,
-                    "channels": channels,
-                    "data": arrow_image["data"].values.to_numpy().astype(storage_type),
-                }
-
-                plot.frame = np.reshape(
-                    image["data"], (image["height"], image["width"], image["channels"])
+                plot.frame = (
+                    storage.to_numpy()
+                    .astype(storage_type)
+                    .reshape((height, width, channels))
+                    .copy()  # Copy So that we can add annotation on the image
                 )
 
                 plot_frame(plot)

node-hub/opencv-video-capture/opencv_video_capture/main.py

@@ -1,6 +1,6 @@
 import os
 import argparse
-from cv2 import cv2  # Importing cv2 this way remove error warnings``
+import cv2
 
 import numpy as np
 import pyarrow as pa
@@ -105,14 +105,14 @@ def main():
                 if image_width is not None and image_height is not None:
                     frame = cv2.resize(frame, (image_width, image_height))
 
-                image = {
-                    "width": np.uint32(frame.shape[1]),
-                    "height": np.uint32(frame.shape[0]),
-                    "encoding": "bgr8",
-                    "data": frame.ravel(),
-                }
+                storage = pa.array(frame.ravel())
 
-                node.send_output("image", pa.array([image]), event["metadata"])
+                metadata = event["metadata"]
+                metadata["width"] = int(frame.shape[1])
+                metadata["height"] = int(frame.shape[0])
+                metadata["encoding"] = "bgr8"
+
+                node.send_output("image", storage, metadata)
 
         elif event_type == "ERROR":
             raise RuntimeError(event["error"])

node-hub/ultralytics-yolo/ultralytics_yolo/main.py

@@ -45,8 +45,11 @@ def main():
             event_id = event["id"]
 
             if event_id == "image":
-                arrow_image = event["value"][0]
-                encoding = arrow_image["encoding"].as_py()
+                storage = event["value"]
+                metadata = event["metadata"]
+                encoding = metadata["encoding"]
+                width = metadata["width"]
+                height = metadata["height"]
 
                 if encoding == "bgr8":
                     channels = 3
@@ -54,18 +57,11 @@ def main():
                 else:
                     raise RuntimeError(f"Unsupported image encoding: {encoding}")
 
-                image = {
-                    "width": np.uint32(arrow_image["width"].as_py()),
-                    "height": np.uint32(arrow_image["height"].as_py()),
-                    "encoding": encoding,
-                    "channels": channels,
-                    "data": arrow_image["data"].values.to_numpy().astype(storage_type),
-                }
-
-                frame = image["data"].reshape(
-                    (image["height"], image["width"], image["channels"])
+                frame = (
+                    storage.to_numpy()
+                    .astype(storage_type)
+                    .reshape((height, width, channels))
                 )
-
                 if encoding == "bgr8":
                     frame = frame[:, :, ::-1]  # OpenCV image (BGR to RGB)
                 else: