tjuskyzhang

Results 22 comments of tjuskyzhang

The input image can be any size. At the preprocessing stage, the input image will be resized into the input size of the network according to your's yolov4-tiny.cfg.

> In the README.md, it is mentioned that the inference speed on TX2 is around 10-11ms. Is this speed measured around the function doInference here: https://github.com/tjuskyzhang/yolov4-tiny-tensorrt/blob/bc49483e49e4de698fd88b878799b8b0a979e88f/yolov4-tiny.cpp#L496. > > Or around...

> Hi, > > Firstly, thanks for this great piece of work! > > I would like to train YOLOv4-Tiny on my custom dataset and the use-case is to run...

Try to set the MAX_OUTPUT_BBOX_COUNT to a large value like 99999999. And nms method maybe different from the pytorch.

please try https://github.com/WongKinYiu/PyTorch_YOLOv4/tree/u3_preview with yolov4-tiny.pt

A yolov4-csp/p5/p6/p7-tensorrt project https://github.com/tjuskyzhang/Scaled-YOLOv4-TensorRT

I have made a mobilenetv1-ssd-tensorrt project based on this project. https://github.com/tjuskyzhang/mobilenetv1-ssd-tensorrt

> @tosonw > 请问yolov4-tiny 您可以使用了么? 我实现了一版yolov4-tiny-tensorrt https://github.com/tjuskyzhang/yolov4-tiny-tensorrt

你看看是不是这个情况#17 > 您好!我使用您的方法[yolov4-tiny-tensorrt](https://github.com/tjuskyzhang/Scaled-YOLOv4-TensorRT/tree/master/yolov4-tiny-tensorrt#yolov4-tiny-tensorrt) 、您的百度网盘里的pt 和 cfg 文件,可以生成 wts、engine,执行 ./yolov4-tiny -d ../samples 命令后生成的图片正常,**每个目标上只有一个框**。 但当我使用训练好的[WongKinYiu/PyTorch_YOLOv4-tiny](https://github.com/WongKinYiu/PyTorch_YOLOv4/tree/master)时,使用我的 pt和cfg文件生成wts、engine后,对我自己的部分数据集图片进行推理时,生成的图片上**每个目标会有多个框**,就像这样 [picture](https://github.com/happyboy688/picture/blob/main/_1.jpeg)。我该如何做? 期待您的回复!

> > 你看看是不是这个情况#17 > > > 您好!我使用您的方法[yolov4-tiny-tensorrt](https://github.com/tjuskyzhang/Scaled-YOLOv4-TensorRT/tree/master/yolov4-tiny-tensorrt#yolov4-tiny-tensorrt) 、您的百度网盘里的pt 和 cfg 文件,可以生成 wts、engine,执行 ./yolov4-tiny -d ../samples 命令后生成的图片正常,**每个目标上只有一个框**。 但当我使用训练好的[WongKinYiu/PyTorch_YOLOv4-tiny](https://github.com/WongKinYiu/PyTorch_YOLOv4/tree/master)时,使用我的 pt和cfg文件生成wts、engine后,对我自己的部分数据集图片进行推理时,生成的图片上**每个目标会有多个框**,就像这样 [picture](https://github.com/happyboy688/picture/blob/main/_1.jpeg)。我该如何做? 期待您的回复! > > 感谢回复!我尝试过那个方法,但是不行。这是我的 cfg、yololayer.h 文件和权重文件 [yolov4-tiny+tensorrt.zip](https://github.com/tjuskyzhang/Scaled-YOLOv4-TensorRT/files/10056686/yolov4-tiny%2Btensorrt.zip)。 > > 这是 yololayer.h...