Adding OAKD support to Donkeycar #1162
Conversation
* Add oak_d camera * Add to config * Add rgb to config
* Add oak_d camera * Add to config * Add rgb to config * Add test to manage.py * Add small fix * Update inputs for OAK-D * fixes to previous issues --------- Co-authored-by: Moises Lopez <[email protected]> Co-authored-by: Moisés López <[email protected]>
|
The donkeydocs github has the docs. You can add documentation to the cameras page https://github.com/autorope/donkeydocs/blob/master/docs/parts/cameras.md |
mdlopezme
force-pushed
the
main
branch
2 times, most recently
from
0ac3693
to
b612919
Compare
donkeycar/parts/oak_d_camera.py
Outdated
| from collections import deque | ||
| import numpy as np | ||
| try: | ||
| import depthai as dai |
There was a problem hiding this comment.
This should go into the RPi or Nano (wherever this is expected to work) section of setup.cfg.
donkeycar/parts/oak_d_camera.py
Outdated
|
|
||
|
|
||
| logger = logging.getLogger(__name__) | ||
| logger.setLevel(logging.INFO) |
There was a problem hiding this comment.
Please remove - logging should not be done on a file / module basis but is passed through in manage.py from the environment or command line instead.
donkeycar/parts/oak_d_camera.py
Outdated
|
|
||
| # Create pipeline | ||
| self.pipeline = dai.Pipeline() | ||
| # self.pipeline.setCameraTuningBlobPath('/tuning_color_ov9782_wide_fov.bin') |
There was a problem hiding this comment.
Pls remove commented code.
donkeycar/parts/oak_d_camera.py
Outdated
|
|
||
| if rgb_apply_cropping: | ||
| camera.setSensorCrop(rgb_sensor_crop_x, rgb_sensor_crop_y) # When croping to keep only smaller video | ||
|
|
There was a problem hiding this comment.
unnecessary blank line
| camera.setSensorCrop(rgb_sensor_crop_x, rgb_sensor_crop_y) # When croping to keep only smaller video | ||
|
|
||
| camera.setVideoSize(rgb_video_size) # Desired video size = ispscale result or smaller if croping | ||
|
|
There was a problem hiding this comment.
additional blank line (pep-8 only wants one within a function).
donkeycar/parts/oak_d_camera.py
Outdated
| self.roi_distances.append(int(coords.y)) | ||
| self.roi_distances.append(int(coords.z)) | ||
|
|
||
| # return self.frame |
| xinSpatialCalcConfig.out.link(spatialLocationCalculator.inputConfig) | ||
|
|
||
|
|
||
| def run(self): |
There was a problem hiding this comment.
This is not correct. The run method should return the same data as run_threaded if the part is called as a non-threaded part. However, if you designed the part to be threaded, then it is fine to not have the run method at all. The update method runs in a separate thread and in general updates member variables which are then used by the run_threaded method. So in case you want to implement the run method as well, it would sense to do it by filling these members accordingly and then returning them the same way as run_threaded, such that everything happens within the run function itself.
| def update(self): | ||
| # Keep looping infinitely until the thread is stopped | ||
| while self.on: | ||
| self.run() |
There was a problem hiding this comment.
This is not how you would generally do it.
donkeycar/parts/oak_d_camera.py
Outdated
| self.device = None | ||
| self.queue = None | ||
| self.pipeline = None | ||
|
|
There was a problem hiding this comment.
Indicates missing EOL.
| @@ -70,7 +70,7 @@ def write_record(self, record=None): | |||
| elif input_type == 'gray16_array': | |||
| # save np.uint16 as a 16bit png | |||
| image = Image.fromarray(np.uint16(value)) | |||
| name = Tub._image_file_name(self.manifest.current_index, key, ext='.png') | |||
| name = Tub._image_file_name(self.manifest.current_index, key, extension='.png') | |||
There was a problem hiding this comment.
Thanks for fixing this, this is indeed a bug!
|
Dear @DocGarbanzo, sorry we've been working from our organization's fork and did not realize we've been receiving many help from you and the |
Still pending testing from different oak cameras. ( We have OakD Pro Wide, OakD Lite, OakD LR)
Is there an specific place were we should add documentation to use this parts?
From Triton AI (Specifically @raymondsong00 's effort)