**• Hardware Platform (Jetson / GPU) XAVIER NX
**• DeepStream Version 5.1
**• JetPack Version (valid for Jetson only) 4.4
**• TensorRT Version 7.1
• NVIDIA GPU Driver Version (valid for GPU only)
I want to integrate the paddleOCR in the jetson to get realtime recognize text,now i have export the paddleOCR detect model and recognize model as onnx model.
I set the det_model as the pgie and the rec_model as sgie.
i refer the demo of deepstream-ssd-parser(python) and deepstream-infer-tensor-meta-test (C++) to learn how to pass the resutl to the secondary,but i can’t get the the sgie resutl, the obj_meta.obj_user_meta_list is None
l_obj=frame_meta.obj_meta_list
while l_obj is not None:
try:
# Casting l_obj.data to pyds.NvDsObjectMeta
obj_meta=pyds.NvDsObjectMeta.cast(l_obj.data)
# tensor_meta = pyds.NvDsInferTensorMeta.cast(obj_meta.obj_meta_data)
print('class_id==',obj_meta.class_id)
print('unique_component_id==',obj_meta.unique_component_id)
l_user = obj_meta.obj_user_meta_list
print ('l_user==',l_user)
while l_user is not None:
try:
# Casting l_obj.data to pyds.NvDsObjectMeta
user_meta=pyds.NvDsUserMeta.cast(l_user.data)
tensor_meta = pyds.NvDsInferTensorMeta.cast(user_meta.user_meta_data)
print(tensor_meta.unique_id)
except StopIteration:
break
try:
l_user = l_user.next
except StopIteration:
break
the l_user is none , i got these out print in the console:
class_id== 1
unique_component_id== 1
l_user== None
my pgie_config.txt as following
[property]
gpu-id=0
net-scale-factor=0.0039215697906911373
model-engine-file=./model/det.engine
#model-color-format=2
#force-implicit-batch-dim=1
infer-dims=3;640;640
#batch-size=1
process-mode=1
network-mode=1
num-detected-classes=1
interval=1
gie-unique-id=1
network-type=100
output-tensor-meta=1
my sgie_config.txt as following
[property]
gpu-id=0
net-scale-factor=0.0039215697906911373
#net-scale-factor=1
#force-implicit-batch-dim=1
model-file=./rec_model.onnx
model-engine-file=./model/rec.engine
gie-unique-id=2
operate-on-gie-id=1
operate-on-class-ids=1
model-color-format=1
infer-dims=3;32;100
batch-size=1
process-mode=2
network-mode=1
interval=0
network-type=100
output-tensor-meta=1
I have found this topic,my problem is similar to this. but these is change to c++, does python can sovle?
https://forums.developer.nvidia.com/t/secondary-inference-using-nvinferserver-after-deepstream-ssd-parser/181773
Anyone can help me? thanks very much