The camera subscription is mandatory for the fusion module. However, you can always fuse the sl.Bodies yourself, without the fusion module.
Hi @alassagne ,
Iām attempting to use the ānetwork mode on USBā setup, but cannot get it to work and donāt see much doc around it. I wrote a simple prototype without multi-processing but things arenāt work. Would you please help to see if Iāve missed anything?
Python:
import pyzed.sl as sl
import os
import time
import cv2
config_path = './svo/fusion-single-network.json'
def main():
fusion_configurations = sl.read_fusion_configuration_file(config_path, sl.COORDINATE_SYSTEM.RIGHT_HANDED_Y_UP, sl.UNIT.METER)
init_fusion_parameters = sl.InitFusionParameters()
body_tracking_fusion_params = sl.BodyTrackingFusionParameters()
body_tracking_runtime_parameters = sl.BodyTrackingRuntimeParameters()
rt = sl.BodyTrackingFusionRuntimeParameters()
fusion = sl.Fusion()
fusion.init(init_fusion_parameters)
fusion.enable_body_tracking(body_tracking_fusion_params)
bodies = {}
mats = {}
cameras = {}
fused = sl.Bodies()
for conf in fusion_configurations:
sn = conf.serial_number
mats[sn] = sl.Mat()
bodies[sn] = sl.Bodies()
svopath = os.path.join(os.path.dirname(config_path), './SVO_SN' + str(sn) + '.svo')
init_params = sl.InitParameters(
camera_resolution = sl.RESOLUTION.VGA,
camera_fps = 15,
svo_real_time_mode = True,
depth_mode = sl.DEPTH_MODE.QUALITY,
coordinate_units = sl.UNIT.METER,
coordinate_system = sl.COORDINATE_SYSTEM.RIGHT_HANDED_Y_UP,
open_timeout_sec = 5, # default = 5
)
init_params.set_from_svo_file(svopath) #Set init parameter to run from the .svo
body_tracking_parameters = sl.BodyTrackingParameters()
body_tracking_parameters.detection_model = sl.BODY_TRACKING_MODEL.HUMAN_BODY_FAST
body_tracking_parameters.body_format = sl.BODY_FORMAT.BODY_38
camera = sl.Camera()
status = camera.open(init_params)
print('camera.open() result:', status)
status = camera.enable_body_tracking(body_tracking_parameters)
print("camera.enable_body_tracking:", status)
communication_parameters = sl.CommunicationParameters()
# communication_parameters.set_for_shared_memory()
communication_parameters.set_for_local_network(conf.communication_parameters.port, conf.communication_parameters.ip_address)
status = camera.start_publishing(communication_parameters)
print('camera.start_publishing() result: ', status)
uuid = sl.CameraIdentifier()
uuid.serial_number = conf.serial_number
status = fusion.subscribe(uuid, conf.communication_parameters, conf.pose)
if status != sl.FUSION_ERROR_CODE.SUCCESS:
print("-- [error] Unable to subscribe to", uuid.serial_number, status)
cameras[sn] = camera
cv2.namedWindow('debug_' + str(sn))
time.sleep(1)
time.sleep(1)
while True:
for sn in cameras:
camera = cameras[sn]
if camera.grab() == sl.ERROR_CODE.SUCCESS:
camera.retrieve_bodies(bodies[sn], body_tracking_runtime_parameters)
camera.retrieve_image(mats[sn], sl.VIEW.LEFT)
cvImage = mats[sn].get_data()
resizedImg = cv2.resize(cvImage, (640, 360))
cv2.imshow('debug_' + str(sn), resizedImg)
fuse_status = fusion.process()
if fuse_status == sl.FUSION_ERROR_CODE.SUCCESS:
fusion.retrieve_bodies(fused, rt)
out = []
for body_data in fused.body_list:
body_out = {}
body_out['id'] = body_data.id
body_out["status"] = str(body_data.tracking_state)
out.append(body_out)
print('fused bodies ', out)
else:
print('fuse failed...', fuse_status)
if cv2.waitKey(10) & 0xFF == ord('q'):
break
main()
fusion config file:
{
"20832444": {
"input": {
"fusion": {
"type": "LOCAL_NETWORK",
"configuration": {
"ip": "192.168.1.100",
"port": 30000
}
},
"zed": {
"configuration": "20832444",
"type": "USB_SERIAL"
}
},
"world": {
"rotation": [
0.0,
0.0,
0.0
],
"translation": [
0.0,
0.0,
0.0
]
}
}
}
Thanks a lot!
I still see USB_SERIAL in your fusion configuration. What if you change it ?
Hi @alassagne would you mind telling me what exactly should I put in there? Are there any working example that I can follow? I did try using SVO_FILE but the error is the same.
{
"20832444": {
"input": {
"fusion": {
"type": "LOCAL_NETWORK",
"configuration": {
"ip": "192.168.1.100",
"port": 30000
}
},
"zed": {
"configuration": "C:\\Users\\user\\source\\repos\\itunnel\\itunnel-depth\\svo\\SVO_SN20832444.svo",
"type": "SVO_FILE"
}
},
"world": {
"rotation": [
0.0,
0.0,
0.0
],
"translation": [
0.0,
0.0,
0.0
]
}
}
}
And the other parameters seems even more irrelevant?
Finally sorted it out.
- The fusion file should be something like this:
"26601708": {
"input": {
"zed": {
"configuration": "C:\\Users\\user\\source\\repos\\itunnel\\itunnel-depth\\svo\\SVO_SN26601708.svo",
"type": "SVO_FILE"
},
"fusion": {
"type": "LOCAL_NETWORK",
"configuration": {
"ip": "127.0.0.1",
"port": 30002
}
}
},
"world": {
"rotation": [
0,
0,
0
],
"translation": [
0,
0,
0
]
}
},
...
- The cameras must be up and running and publishing BEFORE running fusion.subscribe()
Can you tell me how you work with point cloud cropping?
Thanks a lot.
Each 3D point in the point cloud corresponds to a 2D pixel. You can run the convertion using these formula:
With that info, you can either construct the 3D point cloud yourself using only the relevant 2D pixels; or you pass a 2D mask as texture to a glsl shader, and from there decide whether to render a points base on the mask pixel.
1 Like