Uff to TensorRT engine Conversion : uff parsing returns false

I have converted frozen graph to uff but not able to parse the uff file with network, I am getting parser.parse “false” please guide.

#Settings

FROZEN_GDEF_PATH =‘frozen_inference_graph.pb’ # ADJUST
ENGINE_PATH = ‘./engine.plan’ # ADJUST
INPUT_NODE = ‘image_tensor’ # ADJUST
INPUT_SIZE = [1,3, 300, 300] # ADJUST
MAX_BATCH_SIZE = 1 # ADJUST
MAX_WORKSPACE = 1 << 32 # ADJUST
DATA_TYPE = trt.float16 # ADJUST # float16 | float32

Create TRT model builder

trt_logger = trt.Logger(trt.Logger.INFO)
builder = trt.Builder(trt_logger)
builder.max_batch_size = MAX_BATCH_SIZE
builder.max_workspace_size = MAX_WORKSPACE
builder.fp16_mode = (DATA_TYPE == trt.float16)

#Create UFF parser¶
parser = trt.UffParser()
parser.register_input(“image_tensor”, [1,3,300,300],trt.tensorrt.UffInputOrder.NCHW)
parser.register_output(“detection_boxes”)
parser.register_output(“detection_scores”)
parser.register_output(“detection_multiclass_scores”)
parser.register_output(“detection_classes”)
parser.register_output(“num_detections”)
parser.register_output(“raw_detection_boxes”)
parser.register_output(“raw_detection_scores”)
out:True

#Parse UFF graph
uff_model= “/home/pranav/Downloads/My_exp_graph-20200614T112058Z-001/My_exp_graph/frozen_inference_graph.uff”
network = builder.create_network()
parser.parse(uff_model, network)
out: False

Hi,

Did you ever fix this issue? I am having the exact same problem. My parser returns 0 and I have no idea what’s causing it.