Merge pull request #912 from luxonis/develop

1.9.2 release

.gitignore

@@ -59,6 +59,7 @@ resources/*.json
 # Virtual environment
 virtualenv/
 venv/
+.venv/
 
 # DepthAI recordings
 recordings/

.gitmodules

@@ -0,0 +1,3 @@
+[submodule "depthai_sdk/src/depthai_sdk/components/integrations/depthai_pipeline_graph"]
+	path = depthai_sdk/src/depthai_sdk/components/integrations/depthai_pipeline_graph
+	url = https://github.com/luxonis/depthai_pipeline_graph

depthai_sdk/docs/source/conf.py

@@ -22,7 +22,7 @@
 author = 'Luxonis'
 
 # The full version, including alpha/beta/rc tags
-release = '1.0.0'
+release = '1.9.2'
 
 
 # -- General configuration ---------------------------------------------------

depthai_sdk/docs/source/features/recording.rst

@@ -71,7 +71,7 @@ stream and mono streams. You can open the rosbag with the `RealSense Viewer <htt
 -----------------
 
 An alternative to Rosbags are `mcap files <https://github.com/foxglove/mcap>`__ which can be viewed with `Foxglove studio <https://foxglove.dev/>`__.
-You can find `MCAP recording example here <https://github.com/luxonis/depthai/blob/main/depthai_sdk/examples/recording/mcap-record.py>`__.
+You can find `MCAP recording example here <https://github.com/luxonis/depthai/blob/main/depthai_sdk/examples/recording/mcap_record.py>`__.
 Currently supported streams:
 
 - MJPEG encoded color/left/right/disparity. Lossless MJPEG/H264/H265 aren't supported by Foxglove Studio.

depthai_sdk/examples/CameraComponent/cam_ffc.py

@@ -0,0 +1,10 @@
+from depthai_sdk import OakCamera
+
+with OakCamera() as oak:
+    cama = oak.create_camera('cama,c', resolution='1200p')
+    camb = oak.create_camera('camb,c', resolution='1200p')
+    camc = oak.create_camera('camc,c', resolution='1200p')
+    # stereo = oak.create_stereo(left=left, right=right)
+
+    oak.visualize([cama, camb,camc], fps=True, scale=2/3)
+    oak.start(blocking=True)

depthai_sdk/examples/CameraComponent/rgb_mono_preview.py

@@ -5,4 +5,4 @@
     left = oak.create_camera('left')
     right = oak.create_camera('right')
     oak.visualize([color, left, right], fps=True)
-    oak.start(blocking=True)
+    oak.start(blocking=True)

depthai_sdk/examples/NNComponent/deeplabv3_person.py

@@ -1,35 +1,11 @@
-import cv2
-import numpy as np
-
-from depthai_sdk import OakCamera, DetectionPacket, Visualizer
-
-NN_WIDTH, NN_HEIGHT = 513, 513
-
-
-def process_mask(output_tensor):
-    class_colors = [[0, 0, 0], [0, 255, 0]]
-    class_colors = np.asarray(class_colors, dtype=np.uint8)
-    output = output_tensor.reshape(NN_WIDTH, NN_HEIGHT)
-    output_colors = np.take(class_colors, output, axis=0)
-    return output_colors
-
-
-def callback(packet: DetectionPacket, visualizer: Visualizer):
-    frame = packet.frame
-    mask = packet.img_detections.mask
-
-    output_colors = process_mask(mask)
-    output_colors = cv2.resize(output_colors, (frame.shape[1], frame.shape[0]))
-
-    frame = cv2.addWeighted(frame, 1, output_colors, 0.2, 0)
-    cv2.imshow('DeepLabV3 person segmentation', frame)
+from depthai_sdk import OakCamera
 
 
 with OakCamera() as oak:
     color = oak.create_camera('color', resolution='1080p')
 
     nn = oak.create_nn('deeplabv3_person', color)
-    nn.config_nn(resize_mode='stretch')
+    nn.config_nn(resize_mode='letterbox')
 
-    oak.callback(nn, callback=callback)
+    visualizer = oak.visualize([nn, nn.out.passthrough], fps=True)
     oak.start(blocking=True)

depthai_sdk/examples/NNComponent/mobilenet_encoded.py

@@ -0,0 +1,7 @@
+from depthai_sdk import OakCamera
+
+with OakCamera() as oak:
+    color = oak.create_camera('color', encode='mjpeg', fps=10)
+    nn = oak.create_nn('mobilenet-ssd', color, spatial=True)
+    oak.visualize([nn.out.encoded])
+    oak.start(blocking=True)

depthai_sdk/examples/NNComponent/roboflow_integration.py

@@ -8,8 +8,8 @@
     model_config = {
         'source': 'roboflow', # Specify that we are downloading the model from Roboflow
         'model':'american-sign-language-letters/6',
-        'key':'d2OP8nbhA9rZcWd6G8p1' # Fake API key, replace with your own!
+        'key':'181b0f6e43d59ee5ea421cd77f6d9ea2a4b059f8' # Fake API key, replace with your own!
     }
     nn = oak.create_nn(model_config, color)
     oak.visualize(nn, fps=True)
-    oak.start(blocking=True)
+    oak.start(blocking=True)

depthai_sdk/examples/NNComponent/yolo.py

@@ -3,6 +3,6 @@
 with OakCamera() as oak:
     color = oak.create_camera('color')
     nn = oak.create_nn('yolo-v3-tf', color)
-    oak.visualize([nn, color], scale=2 / 3, fps=True) # 1080P -> 720P
+    oak.visualize([nn, color], scale=2 / 3, fps=True)  # 1080P -> 720P
     # oak.show_graph()
-    oak.start(blocking=True)
+    oak.start(blocking=True)

depthai_sdk/examples/StereoComponent/stereo.py

@@ -1,17 +1,17 @@
 import cv2
 
 from depthai_sdk import OakCamera
+from depthai_sdk.components.stereo_component import WLSLevel
 from depthai_sdk.visualize.configs import StereoColor
 
 with OakCamera() as oak:
-    stereo = oak.create_stereo('400p', fps=30)
-    stereo.configure_postprocessing(
-        colorize=StereoColor.RGB,
+    stereo = oak.create_stereo('800p', fps=30)
+    stereo.config_postprocessing(
+        colorize=StereoColor.GRAY,
         colormap=cv2.COLORMAP_BONE,
         wls_filter=True,
-        wls_lambda=8000,
-        wls_sigma=1.5
+        wls_level=WLSLevel.HIGH
     )
 
-    oak.visualize(stereo.out.depth)
+    oak.visualize(stereo.out.disparity)
     oak.start(blocking=True)

depthai_sdk/examples/StereoComponent/stereo_encoded.py

@@ -0,0 +1,9 @@
+import cv2
+
+from depthai_sdk import OakCamera
+from depthai_sdk.visualize.configs import StereoColor
+
+with OakCamera() as oak:
+    stereo = oak.create_stereo('400p', fps=30, encode='h264')
+    oak.visualize(stereo.out.encoded)
+    oak.start(blocking=True)

depthai_sdk/examples/mixed/sync_multiple_outputs.py

@@ -3,9 +3,9 @@
 from depthai_sdk import OakCamera
 
 with OakCamera() as oak:
-    color = oak.create_camera('color', encode='h264')
-    nn = oak.create_nn('mobilenet-ssd', color)
-    nn2 = oak.create_nn('face-detection-retail-0004', color)
+    color = oak.create_camera('color', encode='h264', name='color')
+    nn = oak.create_nn('mobilenet-ssd', color, name='mobilenet')
+    nn2 = oak.create_nn('face-detection-retail-0004', color, name='face-detection')
     # oak.visualize([nn.out.main, nn.out.passthrough])
     # oak.visualize(nn.out.spatials, scale=1 / 2)
     def cb(msgs: Dict):

depthai_sdk/examples/recording/encode.py

@@ -1,15 +1,15 @@
 from depthai_sdk import OakCamera, RecordType
 
 with OakCamera() as oak:
-    color = oak.create_camera('color', resolution='1080P', fps=20, encode='H265')
-    left = oak.create_camera('left', resolution='800p', fps=20, encode='H265')
-    right = oak.create_camera('right', resolution='800p', fps=20, encode='H265')
+    color = oak.create_camera('color', resolution='1080P', fps=10, encode='H265')
+    left = oak.create_camera('left', resolution='800p', fps=10, encode='H265')
+    right = oak.create_camera('right', resolution='800p', fps=10, encode='H265')
 
     stereo = oak.create_stereo(left=left, right=right)
     nn = oak.create_nn('mobilenet-ssd', color, spatial=stereo)
 
     # Sync & save all (encoded) streams
-    oak.record([color.out.encoded, left.out.encoded, right.out.encoded], './', RecordType.VIDEO)
-    oak.visualize([nn])
+    oak.record([color.out.encoded, left.out.encoded, right.out.encoded], './record', RecordType.VIDEO)
+    oak.visualize([color.out.encoded], fps=True)
 
-    oak.start(blocking=True)
+    oak.start(blocking=True)

depthai_sdk/examples/recording/mcap-record.py → depthai_sdk/examples/recording/mcap_record.py

@@ -2,7 +2,7 @@
 
 with OakCamera() as oak:
     color = oak.create_camera('color', resolution='1080P', fps=30, encode='MJPEG')
-    color.config_color_camera(ispScale=(2,3)) # 720P
+    color.config_color_camera(isp_scale=(2, 3)) # 720P
     left = oak.create_camera('left', resolution='400p', fps=30)
     right = oak.create_camera('right', resolution='400p', fps=30)
     stereo = oak.create_stereo(left=left, right=right)

depthai_sdk/examples/recording/stereo_record.py

@@ -1,14 +1,13 @@
 import cv2
-import depthai
 
 from depthai_sdk import OakCamera
 from depthai_sdk.visualize.configs import StereoColor
 
 with OakCamera() as oak:
-    color = oak.create_camera('color', resolution='1080p', fps=5)
-    stereo = oak.create_stereo('400p', fps=5)
+    color = oak.create_camera('color', resolution='1080p', fps=30)
+    stereo = oak.create_stereo('400p', fps=30)
 
-    stereo.configure_postprocessing(
+    stereo.config_postprocessing(
         colorize=StereoColor.RGB,
         colormap=cv2.COLORMAP_JET,
         wls_filter=True,
@@ -18,9 +17,9 @@
 
     # Record RGB and disparity to records folder
     # Record doesn't work with visualize so the config is ignored
-    oak.record([color.out.main, stereo.out.disparity], 'records')
+    # oak.record([color.out.main, stereo.out.disparity], 'records')
 
     # Record depth only
-    # oak.visualize(stereo.out.depth, record='depth.avi')
+    oak.visualize(stereo.out.disparity, record_path='disparity.avi')
 
     oak.start(blocking=True)

depthai_sdk/examples/replay/people-tracker.py

@@ -1,5 +1,3 @@
-import cv2
-import depthai as dai
 from depthai_sdk import OakCamera, ResizeMode
 
 with OakCamera(replay="people-tracking-above-02") as oak:

depthai_sdk/requirements.txt

@@ -3,7 +3,9 @@ numpy>=1.21; python_version >= "3.7"
 opencv-contrib-python>4
 blobconverter>=1.2.8
 pytube>=12.1.0
-depthai>=2.18.0
+--extra-index-url https://artifacts.luxonis.com/artifactory/luxonis-python-snapshot-local/
+# This specific commit is needed for the debug mode (oak.show_graph()). TODO: update when depthai has new release
+depthai>=2.20.0.0
 PyTurboJPEG==1.6.4
 marshmallow==3.17.0
 distinctipy

depthai_sdk/setup.py

@@ -3,11 +3,14 @@
 from setuptools import setup
 
 with open('requirements.txt') as f:
-    required = f.readlines()
+    requirements = f.readlines()
+
+install_requires=[requirement for requirement in requirements if '--' not in requirement]
+
 
 setup(
     name='depthai-sdk',
-    version='1.9.1.1',
+    version='1.9.2',
     description='This package provides an abstraction of the DepthAI API library.',
     long_description=io.open("README.md", encoding="utf-8").read(),
     long_description_content_type="text/markdown",
@@ -18,7 +21,7 @@
     license='MIT',
     packages=['depthai_sdk'],
     package_dir={"": "src"},  # https://stackoverflow.com/a/67238346/5494277
-    install_requires=required,
+    install_requires=install_requires,
     include_package_data=True,
     extras_require={
         "visualize": ['PySide2',

depthai_sdk/src/depthai_sdk/classes/nn_config.py

@@ -1,4 +1,5 @@
 from marshmallow import Schema, fields
+
 from depthai_sdk.classes.yolo_config import YoloConfig
 
 

depthai_sdk/src/depthai_sdk/classes/nn_results.py

@@ -19,6 +19,7 @@ class GenericNNOutput:
     def __init__(self, nn_data: NNData):
         self.nn_data = nn_data
 
+
 # First we have Object detection results, which are already standarized with dai.ImgDetections
 
 @dataclass
@@ -59,7 +60,7 @@ class SemanticSegmentation(GenericNNOutput):  # In core, extend from NNData
 
     def __init__(self, nn_data: NNData, mask: List[np.ndarray]):
         super().__init__(nn_data)
-        self.mask: List[np.ndarray] = mask
+        self.mask = mask
 
 
 @dataclass
@@ -69,17 +70,16 @@ class ImgLandmarks(GenericNNOutput):  # In core, extend from NNData
 
     Examples: `human-pose-estimation-0001`, `openpose2`, `facial-landmarks-68`, `landmarks-regression-retail-0009`.
     """
-    landmarks: List[List[Any]]
-    pairs: List[Tuple[int, int]]  # Pairs of landmarks, to draw lines between them
-    colors: List[Tuple[int, int, int]]  # Color for each landmark (eg. both elbows are in the same color)
 
     def __init__(self,
                  nn_data: NNData,
                  landmarks: List[List[Any]] = None,
+                 landmarks_indices: List[List[int]] = None,
                  pairs: List[Tuple[int, int]] = None,
                  colors: List[Tuple[int, int, int]] = None):
         super().__init__(nn_data)
         self.landmarks = landmarks
+        self.landmarks_indices = landmarks_indices
         self.pairs = pairs
         self.colors = colors
 
@@ -89,9 +89,10 @@ class InstanceSegmentation(GenericNNOutput):
     """
     Instance segmentation results, with a mask for each instance.
     """
+    # TODO: Finish this, add example models
     masks: List[np.ndarray]  # 2D np.array for each instance
     labels: List[int]  # Class label for each instance
 
     def __init__(self, nn_data: NNData, masks: List[np.ndarray], labels: List[int]):
+        raise NotImplementedError('Instance segmentation not yet implemented')
         super().__init__(nn_data)
-