MXNet——如何使用训练好的MobileNet1.0

最新推荐文章于 2024-08-09 07:29:42 发布

kaichu2

最新推荐文章于 2024-08-09 07:29:42 发布

阅读量1.3k

点赞数

分类专栏： MXNet计算机视觉文章标签： MXNet——如何使用训练好的MobileNet1.0

本文链接：https://blog.csdn.net/weixin_43687366/article/details/103690523

版权

MXNet计算机视觉专栏收录该内容

7 篇文章 0 订阅

订阅专栏

一、了解MobileNet

推荐：轻量化网络ShuffleNet MobileNet v1/v2 解析

这位知乎博主讲的挺好的，很好理解！不知道的请移步学习！！

下面我们来测试一下分类的效果如何？

二、实际测试一下分类效果

方法一：会自动下载训练好的网络模型，方法二也是一样

from mxnet.gluon.model_zoo import vision
#加载网络参数
mobilenet_v1 = vision.MobileNet()
print(mobilenet_v1.output.weight)

Parameter mobilenet4_dense0_weight (shape=(1000, 0), dtype=float32)

#查看全部的网络参数
print(mobilenet_v1.summary)

<bound method Block.summary of MobileNet(
  (features): HybridSequential(
    (0): Conv2D(None -> 32, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
    (1): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (2): Activation(relu)
    (3): Conv2D(None -> 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
    (4): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (5): Activation(relu)
    (6): Conv2D(None -> 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (7): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (8): Activation(relu)
    (9): Conv2D(None -> 64, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=64, bias=False)
    (10): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (11): Activation(relu)
    (12): Conv2D(None -> 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (13): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (14): Activation(relu)
    (15): Conv2D(None -> 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128, bias=False)
    (16): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (17): Activation(relu)
    (18): Conv2D(None -> 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (19): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (20): Activation(relu)
    (21): Conv2D(None -> 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=128, bias=False)
    (22): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (23): Activation(relu)
    (24): Conv2D(None -> 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (25): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (26): Activation(relu)
    (27): Conv2D(None -> 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=256, bias=False)
    (28): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (29): Activation(relu)
    (30): Conv2D(None -> 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (31): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (32): Activation(relu)
    (33): Conv2D(None -> 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=256, bias=False)
    (34): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (35): Activation(relu)
    (36): Conv2D(None -> 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (37): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (38): Activation(relu)
    (39): Conv2D(None -> 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512, bias=False)
    (40): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (41): Activation(relu)
    (42): Conv2D(None -> 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (43): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (44): Activation(relu)
    (45): Conv2D(None -> 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512, bias=False)
    (46): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (47): Activation(relu)
    (48): Conv2D(None -> 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (49): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (50): Activation(relu)
    (51): Conv2D(None -> 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512, bias=False)
    (52): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (53): Activation(relu)
    (54): Conv2D(None -> 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (55): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (56): Activation(relu)
    (57): Conv2D(None -> 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512, bias=False)
    (58): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (59): Activation(relu)
    (60): Conv2D(None -> 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (61): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (62): Activation(relu)
    (63): Conv2D(None -> 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512, bias=False)
    (64): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (65): Activation(relu)
    (66): Conv2D(None -> 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (67): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (68): Activation(relu)
    (69): Conv2D(None -> 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=512, bias=False)
    (70): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (71): Activation(relu)
    (72): Conv2D(None -> 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (73): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (74): Activation(relu)
    (75): Conv2D(None -> 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=1024, bias=False)
    (76): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (77): Activation(relu)
    (78): Conv2D(None -> 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (79): BatchNorm(axis=1, eps=1e-05, momentum=0.9, fix_gamma=False, use_global_stats=False, in_channels=None)
    (80): Activation(relu)
    (81): GlobalAvgPool2D(size=(1, 1), stride=(1, 1), padding=(0, 0), ceil_mode=True, global_pool=True, pool_type=avg, layout=NCHW)
    (82): Flatten
  )
  (output): Dense(None -> 1000, linear)
)>

#查看第二个conv的参数
print(mobilenet_v1.features[3].weight)
print(mobilenet_v1.features[3].name)

Parameter mobilenet4_conv1_weight (shape=(32, 0, 3, 3), dtype=<class 'numpy.float32'>)
mobilenet4_conv1

下面方法二的查看网络参数的方法也是如此！

方法二：将模型保存在model文件夹中；自动下载的！！！

测试图像：都是我自己标注的名字！

import cv2
img = cv2.imread("test_image/lion.jpg")
cv2.imshow("input image",img)
cv2.waitKey(0)
cv2.destroyAllWindows()

import gluoncv
model_name = 'MobileNet1.0'
net = gluoncv.model_zoo.get_model(model_name,pretrained=True,root="model/")

import mxnet as mx
# or use cv2 import image
img = mx.image.imread("test_image/lion.jpg")
transformed_img = gluoncv.data.transforms.presets.imagenet.transform_eval(img)
pred = net(transformed_img)
# map predicted values to probability by softmax
prob = mx.nd.softmax(pred)[0].asnumpy()
# find the 3 class indices with the highest score
ind = mx.nd.topk(pred, k=3)[0].astype('int').asnumpy().tolist()
print("索引列表为：",ind)
print("\n")
# print the class name and predicted probability
print('The input picture is classified to be')
for i in range(3):
    print('- [%s], with probability %.3f.'%(net.classes[ind[i]], prob[ind[i]]))

索引列表为： [291, 286, 293]


The input picture is classified to be
- [lion], with probability 0.968.
- [cougar], with probability 0.010.
- [cheetah], with probability 0.001.

测试的准确率很高！并且速度也很快！其他图像的测试就不添加了