地址

https://github.com/theAIGuysCode/tensorflow-yolov4-tflite

配置环境

name: yolov4-gpu

dependencies:

• python==3.7
• pip
• matplotlib
• opencv
• cudnn
• cudatoolkit==10.1.243
• pip:
  • tensorflow-gpu==2.3.0rc0
  • opencv-python==4.1.1.26
  • lxml
  • tqdm
  • absl-py
  • easydict
  • pillow

我的tensorflow-gpu2.4.1 ，cuda10.0 能够正常使用。

下载weights文件

https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-tiny.weights

生成模型

python save_model.py --weights ./model/yolov4-tiny.weights --tiny true --output ./checkpoints/yolov4.tf --input_size 416 --model yolov4

进行预测

python detect.py --weights ./checkpoints/yolov4.tf --size 416 --model yolov4 --images kite.jpg --framework trt