模型转为RKNN格式

1.onnx参数说明其中的input_size_list必须是静态参数2.模型参数查看脚本示例输出 输入节点 Name: images Shape: [1, 3, 640, 640] 输出节点 Name: output Shape: [1, 255, 80, 80] Name: 283 Shape: [1, 255, 40, 40] Name: 285 Shape: [1, 255, 20, 20]import onnx # 加载模型 model onnx.load(yolov5s_relu.onnx) # 查看所有输入 print( 输入节点 ) for input in model.graph.input: print(fName: {input.name}) # 获取 shape shape [dim.dim_value if dim.dim_value else dim.dim_param for dim in input.type.tensor_type.shape.dim] print(fShape: {shape}) # print(fDtype: {input.type.tensor_type.elem_type}) # 1FLOAT, 7INT64 # 查看所有输出 print(\n 输出节点 ) for output in model.graph.output: print(fName: {output.name}) shape [dim.dim_value if dim.dim_value else dim.dim_param for dim in output.type.tensor_type.shape.dim] print(fShape: {shape}) # 查看所有中间节点可选 # print(\n 所有节点 ) # for node in model.graph.node: # print(f{node.op_type}: {node.name})3.模型转换脚本from rknn.api import RKNN rknn RKNN(verboseTrue) rknn.config( target_platformrk3568, ) batch_size1 sequence_length512 past_sequence_length512 ret rknn.load_onnx(modelyolov5s_relu.onnx) if ret ! 0: print(Load model failed!) exit(ret) print(done) ret rknn.build(do_quantizationFalse) if ret ! 0: print(Load model failed!) exit(ret) print(done) ret rknn.export_rknn(export_path./rknn.rknn) if ret ! 0: print(Load model failed!) exit(ret) print(done)