GeekAlexis / FastMOT

High-performance multiple object tracking based on YOLO, Deep SORT, and KLT 🚀
MIT License
1.12k stars 256 forks source link

Failed to parse the ONNX file #232

Closed 43998213 closed 2 years ago

43998213 commented 2 years ago

/home/clb/github/FastMOT-2.0.0/fastmot/models/osnet_x0_25_msmt17.trt 2021-12-23 10:38:38 [ INFO] Building engine with batch size: 16 2021-12-23 10:38:38 [ INFO] This may take a while... [TensorRT] ERROR: (Unnamed Layer* 486) [Shuffle]: at most one dimension may be inferred 2021-12-23 10:38:38 [CRITICAL] Failed to parse the ONNX file 2021-12-23 10:38:38 [ ERROR] In node -1 (scaleHelper): UNSUPPORTED_NODE: Assertion failed: dims.nbDims == 4 || dims.nbDims == 5 Traceback (most recent call last): File "app.py", line 101, in main() File "app.py", line 53, in main mot = fastmot.MOT(config.resize_to, vars(config.mot_cfg), draw=draw) File "/home/clb/github/FastMOT-2.0.0/fastmot/mot.py", line 88, in init self.extractor = FeatureExtractor(vars(feature_extractor_cfg)) File "/home/clb/github/FastMOT-2.0.0/fastmot/feature_extractor.py", line 28, in init self.backend = TRTInference(self.model, self.batch_size) File "/home/clb/github/FastMOT-2.0.0/fastmot/utils/inference.py", line 63, in init raise RuntimeError('Unable to load the engine file')

jetson agx

import tensorrt as rt rt.version '7.2.0.14'

onnx==1.4.1

help thanks