YangShuai

> @armaab 不是GPU，现在原作者更新了另外一种识别方法， CNN+RNN+CTC，其中RNN用的是GRU。 > 不过现在我已经自己写了训练，还是谢谢你。:) 你好，我想问一下你是用的gru训练还是用的lstm呀？在车牌序列识别中，你觉得gru和lstm哪个更合适呢

same error. convert ultralytics/yolov5 to int8 with tensorrt8 and encountered this error. convert is ok with tensorrt7. any solution?

> same error CUDA10.1、CUDNN7.5、tensorrt install by pip wheel hi, do you solved this error?

> > > > google的门牌号识别的论文里说，第一个softmax用于预测序列长度，其余的softmax用于预测位置和字符。照这样来说，最后不应该用8个softmax吗？车牌号长度一般为7，第一个softmax用于预测车牌号序列长度，剩下7个预测位置和字符。 > > 请问google门牌号码识别的论文在哪里有 https://arxiv.org/abs/1312.6082

收到，感谢ღ( ´･ᴗ･` )比心

> > > ``` > > > I have found it, thanks > > > ``` > > > > > > hello，i dont know how to output the province....

> Use [deepstream support](https://github.com/grimoire/mmdetection-to-tensorrt/blob/master/docs/getting_started.md#deepstream-support) if you want to do stream inference. Or use c++ api like [official one](https://docs.nvidia.com/deeplearning/tensorrt/developer-guide/index.html#perform_inference_c). > Preprocess and postprocess are not in the engine. Should be implemented...

> > > @ShaneYS The only **postprocess** you need to do is > > ``` > bboxes = bboxes/scale_factor > ``` 谢谢 还是有点问题。 在导出onnx或者TensorRT时，什么类型的操作会被导出，什么类型的操作不会被导出？ 为什么在很多pytorch转TensorRT的代码中，都说nms操作无法导出，需要自己实现？

> @ShaneYS > [Here](https://docs.nvidia.com/deeplearning/tensorrt/api/python_api/infer/Graph/Layers.html) is the TensorRT supported layers. > Ops like nms does not supported can be implemented as [TensorRT Plugin](https://docs.nvidia.com/deeplearning/tensorrt/developer-guide/index.html#extending). That's why this project need `amirstan_plugin`. Got it,...

> > > Will give a try. > Thank you! hi, have you train your own data? I also want to train segdecnet on my own data( I have images...