Update depthai-core

Author: Erol444

depthai-core

@@ -1,8 +1,9 @@
 Multi-Input Frame Concationation
 ================================
 
-Example concatenates all 3 inputs with a simple custom NN created with PyTorch (`link here <https://github.com/luxonis/depthai-experiments/blob/master/gen2-custom-models/generate_model/pytorch_concat.py>`__, `tutorial here <https://docs.luxonis.com/en/latest/pages/tutorials/creating-custom-nn-models/>`__).
-It uses multiple input :ref:`NeuralNetwork` feature and links all 3 camera streams directly to the NN node.
+Example concatenates all 3 inputs with a simple custom model created with PyTorch (`link here <https://github.com/luxonis/depthai-experiments/blob/master/gen2-custom-models/generate_model/pytorch_concat.py>`__,
+`tutorial here <https://docs.luxonis.com/en/latest/pages/tutorials/creating-custom-nn-models/>`__).
+It uses :ref:`NeuralNetwork`'s multiple input feature and links all 3 camera streams directly to the NeuralNetwork node.
 
 Demo
 ####
@@ -31,7 +32,7 @@ Source code
 
         Also `available on GitHub <https://github.com/luxonis/depthai-core/blob/main/examples/src/concat_multiple_input.cpp>`__
 
-        .. literalinclude:: ../../../../depthai-core/examples/src/concat_multi_input.cpp
+        .. literalinclude:: ../../../../depthai-core/examples/NeuralNetwork/concat_multi_input.cpp
            :language: cpp
            :linenos:
 

docs/source/samples/NeuralNetwork/concat_multi_input.rst

@@ -1,21 +1,21 @@
-Multi-Input Frame Normalization
-===============================
+Frame Normalization
+===================
 
-This example shows how you can normalize a frame before sending it to another NN. Many neural network models
+This example shows how you can normalize a frame before sending it to another neural network. Many neural network models
 require frames with RGB values (pixels) in range between :code:`-0.5` to :code:`0.5`. :ref:`ColorCamera`'s preview outputs
-values between :code:`0` and :code:`255`. Simple custom NN, created with PyTorch (`link here <https://github.com/luxonis/depthai-experiments/blob/master/gen2-custom-models/generate_model/pytorch_normalize.py>`__, `tutorial here <https://docs.luxonis.com/en/latest/pages/tutorials/creating-custom-nn-models/>`__),
-allows users to specify mean and scale factors that will apply for all frame values (pixels).
+values between :code:`0` and :code:`255`. Simple custom model, created with PyTorch (`link here <https://github.com/luxonis/depthai-experiments/blob/master/gen2-custom-models/generate_model/pytorch_normalize.py>`__, `tutorial here <https://docs.luxonis.com/en/latest/pages/tutorials/creating-custom-nn-models/>`__),
+allows users to specify mean and scale factors that will be applied to all frame values (pixels).
 
 .. math::
 
    output = (input - mean) / scale
 
-.. image:: /_static/images/examples/normalize)model.png
+.. image:: /_static/images/examples/normalize_model.png
 
 On the host, values are converted back to :code:`0`-:code:`255`, so they can be displayed by OpenCV.
 
 .. note::
-    This is just a demo, for normalization you should look into OpenVINO's `model optimizer <https://docs.openvinotoolkit.org/latest/openvino_docs_MO_DG_prepare_model_convert_model_Converting_Model_General.html>`__ arguments :code:`--mean_values` and :code:`--scale_values`.
+    This is just a demo, for normalization you should use OpenVINO's `model optimizer <https://docs.openvinotoolkit.org/latest/openvino_docs_MO_DG_prepare_model_convert_model_Converting_Model_General.html>`__ arguments :code:`--mean_values` and :code:`--scale_values`.
 
 Setup
 #####
@@ -39,7 +39,7 @@ Source code
 
         Also `available on GitHub <https://github.com/luxonis/depthai-core/blob/main/examples/src/normalization_multiple_input.cpp>`__
 
-        .. literalinclude:: ../../../../depthai-core/examples/src/normalization_multi_input.cpp
+        .. literalinclude:: ../../../../depthai-core/examples/NeuralNetwork/normalization_multi_input.cpp
            :language: cpp
            :linenos:
 

docs/source/samples/NeuralNetwork/normalization_multi_input.rst

@@ -16,6 +16,7 @@ Code Samples
    ../samples/mixed/*
    ../samples/MobileNet/*
    ../samples/MonoCamera/*
+   ../samples/NeuralNetwork/*
    ../samples/ObjectTracker/*
    ../samples/Script/*
    ../samples/SpatialDetection/*
@@ -95,6 +96,11 @@ are presented with code.
 - :ref:`Mono Camera Control` - Demonstrates how to control the mono camera (crop, exposure, sensitivity) from the host
 - :ref:`Mono Full Resolution Saver` - Saves mono (720P) images to the host (:code:`.png`)
 
+.. rubric:: NeuralNetwork
+
+- :ref:`Multi-Input Frame Concat <Multi-Input Frame Concationation>` - Concat mono/rgb streams on the device with a custom model
+- :ref:`Frame Normalization` - Normalize the frame on the device with a custom model
+
 .. rubric:: ObjectTracker
 
 - :ref:`Object tracker on video` - Performs object tracking from the video

docs/source/tutorials/code_samples.rst

@@ -24,34 +24,26 @@
 camRgb.setInterleaved(False)
 camRgb.setColorOrder(dai.ColorCameraProperties.ColorOrder.BGR)
 
-left = p.create(dai.node.MonoCamera)
-left.setBoardSocket(dai.CameraBoardSocket.LEFT)
-left.setResolution(dai.MonoCameraProperties.SensorResolution.THE_400_P)
-
-# ImageManip for cropping (face detection NN requires input image of 300x300) and to change frame type
-manipLeft = p.create(dai.node.ImageManip)
-manipLeft.initialConfig.setResize(300, 300)
-manipLeft.initialConfig.setFrameType(dai.RawImgFrame.Type.BGR888p)
-left.out.link(manipLeft.inputImage)
-
-right = p.create(dai.node.MonoCamera)
-right.setBoardSocket(dai.CameraBoardSocket.RIGHT)
-right.setResolution(dai.MonoCameraProperties.SensorResolution.THE_400_P)
-
-# ImageManip for cropping (face detection NN requires input image of 300x300) and to change frame type
-manipRight = p.create(dai.node.ImageManip)
-manipRight.initialConfig.setResize(300, 300)
-manipRight.initialConfig.setFrameType(dai.RawImgFrame.Type.BGR888p)
-right.out.link(manipRight.inputImage)
+def create_mono(p, socket):
+    mono = p.create(dai.node.MonoCamera)
+    mono.setBoardSocket(socket)
+    mono.setResolution(dai.MonoCameraProperties.SensorResolution.THE_400_P)
+
+    # ImageManip for cropping (face detection NN requires input image of 300x300) and to change frame type
+    manip = p.create(dai.node.ImageManip)
+    manip.initialConfig.setResize(300, 300)
+    manip.initialConfig.setFrameType(dai.RawImgFrame.Type.BGR888p)
+    mono.out.link(manip.inputImage)
+    return manip.out
 
 # NN that detects faces in the image
 nn = p.createNeuralNetwork()
 nn.setBlobPath(nnPath)
 nn.setNumInferenceThreads(2)
 
-manipLeft.out.link(nn.inputs['img1'])
 camRgb.preview.link(nn.inputs['img2'])
-manipRight.out.link(nn.inputs['img3'])
+create_mono(p, dai.CameraBoardSocket.LEFT).link(nn.inputs['img1'])
+create_mono(p, dai.CameraBoardSocket.RIGHT).link(nn.inputs['img3'])
 
 # Send bouding box from the NN to the host via XLink
 nn_xout = p.createXLinkOut()

examples/NeuralNetwork/concat_multi_input.py

@@ -31,7 +31,7 @@
 
 script = p.create(dai.node.Script)
 script.setScript("""
-# Run script only once
+# Run script only once. We could also send these values from host.
 # Model formula:
 # output = (input - mean) / scale
 
@@ -68,10 +68,11 @@
     shape = (3, SHAPE, SHAPE)
     while True:
         inNn = np.array(qNn.get().getData())
-        # Get back the frame. It's currently normalized to -1.0 - 1.0
+        # Get back the frame. It's currently normalized to -0.5 - 0.5
         frame = inNn.view(np.float16).reshape(shape).transpose(1, 2, 0)
-        # To get original frame back (0-255), we add multiply all frame values (pixels) by 255 and then add 127.5 to them.
+        # To get original frame back (0-255), we add multiply all frame values (pixels) by 255 and then add 127.5 to them
         frame = (frame * 255.0 + 127.5).astype(np.uint8)
+        # Show the initial frame
         cv2.imshow("Original frame", frame)
 
         if cv2.waitKey(1) == ord('q'):