网站建设及发展,网站注册页面跳出怎么做,单位网站建设的目的,wordpress直接上传视频网站吗一直想做一点3D目标检测#xff0c;先来一篇单目3D目标检测Monodle#xff08;基于centernet的#xff09;#xff0c;训练代码参考官方【代码】#xff0c;这里只讲讲如何部署。 模型和完整仿真测试代码#xff0c;放在github上参考链接【模型和完整代码】。
1 模型训练… 一直想做一点3D目标检测先来一篇单目3D目标检测Monodle基于centernet的训练代码参考官方【代码】这里只讲讲如何部署。 模型和完整仿真测试代码放在github上参考链接【模型和完整代码】。
1 模型训练 训练参考官方代码 https://github.com/xinzhuma/monodle
2 导出onnx 如果按照官方代码导出的onnx后处理写起来比较复杂且后处理时耗比较长这里将后处理的部分代码放到模型中。原始官方导出的onnx模型如下图 本示例导出的onnx模型如下图这样导出便于写后处理代码模型后处理整个时耗也比较优 把centernet3d.py 文件拷贝一份命名为export_onnx.py并进行如下修改 export_onnx.py 修改后的完整代码
import os
import cv2
import torch
import torch.nn as nn
import numpy as npfrom lib.backbones import dla
from lib.backbones.dlaup import DLAUp
from lib.backbones.hourglass import get_large_hourglass_net
from lib.backbones.hourglass import load_pretrian_modelclass CenterNet3D(nn.Module):def __init__(self, backbonedla34, neckDLAUp, num_class3, downsample4):CenterNet for monocular 3D object detection.:param backbone: the backbone of pipeline, such as dla34.:param neck: the necks of detection, such as dla_up.:param downsample: the ratio of down sample. [4, 8, 16, 32]:param head_conv: the channels of convolution in head. default: 256assert downsample in [4, 8, 16, 32]super().__init__()self.heads {heatmap: num_class, offset_2d: 2, size_2d: 2, depth: 2, offset_3d: 2, size_3d: 3,heading: 24}self.backbone getattr(dla, backbone)(pretrainedTrue, return_levelsTrue)channels self.backbone.channels # channels list for feature maps generated by backboneself.first_level int(np.log2(downsample))scales [2 ** i for i in range(len(channels[self.first_level:]))]self.neck DLAUp(channels[self.first_level:], scales_listscales) # feature fusion [such as DLAup, FPN]self.heatmapmaxpool nn.MaxPool2d(kernel_size3, stride1, padding1)# initialize the head of pipeline, according to heads setting.for head in self.heads.keys():if head ! heatmap:output_channels self.heads[head]fc nn.Sequential(nn.Conv2d(channels[self.first_level], 256, kernel_size3, padding1, biasTrue),nn.ReLU(inplaceTrue),nn.Conv2d(256, output_channels, kernel_size1, stride1, padding0, biasTrue))else:output_channels self.heads[head]fc nn.Sequential(nn.Conv2d(channels[self.first_level], 256, kernel_size3, padding1, biasTrue),nn.ReLU(inplaceTrue),nn.Conv2d(256, output_channels, kernel_size1, stride1, padding0, biasTrue),nn.Sigmoid())# initializationif heatmap in head:fc[-2].bias.data.fill_(-2.19)else:self.fill_fc_weights(fc)self.__setattr__(head, fc)def forward(self, input):feat self.backbone(input)feat self.neck(feat[self.first_level:])ret {}for head in self.heads:ret[head] self.__getattr__(head)(feat)if head heatmap:heatmapmax self.heatmapmaxpool(ret[head])ret.update({heatmapmax: heatmapmax})return retdef fill_fc_weights(self, layers):for m in layers.modules():if isinstance(m, nn.Conv2d):nn.init.normal_(m.weight, std0.001)if m.bias is not None:nn.init.constant_(m.bias, 0)def export_onnx(model):print( onnx )dummy_input torch.randn(1, 3, 384, 1280)input_names [data]output_names [heatmap, offset_2d, size_2d, depth, offset_3d, size_3d, heading, heatmapmax]torch.onnx.export(model, dummy_input, ./Monodle_epoch_140.onnx, verboseFalse, input_namesinput_names,output_namesoutput_names, opset_version11)print( convert onnx Finished! .... )if __name__ __main__:print(This is main ...)CLASSES [Pedestrian, Car, Cyclist]net CenterNet3D(backbonedla34)checkpoint torch.load(./weights/checkpoint_epoch_140.pth,map_locationcpu)net.load_state_dict(checkpoint[model_state], strictTrue)net.eval()export_onnx(net)input torch.randn((1, 3, 1280, 384))print(input1:, input.shape, input.dtype)output net(input)print(output.keys())print(output[heatmap].shape)print(output[offset_2d].shape)print(output[size_2d].shape)print(output[depth].shape)print(output[offset_3d].shape)print(output[size_3d].shape)print(output[heading].shape)
运行 python export_onnx.py 生成.onn文件。
3 测试效果
官方pytorch 测试效果 onnx 测试效果 特别说明 由于官方代码的2d框是用3d框计算得到的而本博客是直接解码的模型预测出的2d框所以2d框有所出入。
4 onnx、rknn、horizon、tensorRT测试转完整代码 模型和完整仿真测试代码、测试图片参考【模型和完整代码】。
