search

dog-qiuqiu / Yolo-Fastest

:zap: Based on yolo's ultra-lightweight universal target detection algorithm, the calculation amount is only 250mflops, the ncnn model size is only 666kb, the Raspberry Pi 3b can run up to 15fps+, and the mobile terminal can run up to 178fps+
Other
1.98k stars 427 forks source link

onnx convert error #18

Open AIgiraffe opened 3 years ago

AIgiraffe commented 3 years ago

the code of https://github.com/CaoWGG/TensorRT-YOLOv4 can`t converts the yolo-fastest to onnx model. typeerror: buffer is too small for requested array

MuhammadAsadJaved commented 3 years ago

@jasonlove521 Have you resolved this issue? Are you trying to run Nvidia Jetson devices? Xavier NX etc?

AIgiraffe commented 3 years ago

@MuhammadAsadJaved I train the model by using yolov5-torch, then convert to tensort engine. There is some problem with trt-engine. The mAP and inference speed of trt-engine`s are verlt low.

MuhammadAsadJaved commented 3 years ago

Have you tried to convert the Yolo-Fastest model which the provided in the repo?

AIgiraffe commented 3 years ago

@MuhammadAsadJaved yes. i have done, but that repo can`t convert the Yolo-Fastest to trt-engine,because Depthwise Convolution　is used in the Yolo-Fastest, which occur error. i try to modify the code of the provided repo so that it can work, but the performance of trt_engine is not expected with low AP and slow speed.