SSD+FPN+resnet修改输入图像尺寸

最新推荐文章于 2024-06-29 21:30:16 发布

深度学习扛把子

最新推荐文章于 2024-06-29 21:30:16 发布

阅读量1.7k

点赞数

分类专栏：深度学习知识框架与python

本文链接：https://blog.csdn.net/qq_16792139/article/details/111881229

版权

深度学习知识框架与python 专栏收录该内容

38 篇文章 0 订阅

订阅专栏

来看一下ssd300与ssd512的cfg文件对比
# SSD300 CONFIGS
# voc = {
#     #'num_classes': 21,
#     'num_classes': 7,
#     'lr_steps': (80000, 100000, 120000),
#     'max_iter': 180000,
#     #'max_iter': 10000,
#     'feature_maps': [38, 19, 10, 5, 3, 1],
#     'min_dim': 300,
#     'steps': [8, 16, 32, 64, 100, 300],
#     'min_sizes': [30, 60, 111, 162, 213, 264],
#     'max_sizes': [60, 111, 162, 213, 264, 315],
#     'aspect_ratios': [[2], [2, 3], [2, 3], [2, 3], [2], [2]],
#     'variance': [0.1, 0.2],
#     'clip': True,
#     'name': 'VOC',
# }
# SSD512 CONFIGS
voc = {
    #'num_classes': 21,
    'num_classes': 7,
    'lr_steps': (80000, 100000, 120000),
    'max_iter': 180000,
    #'max_iter': 10000,
    'feature_maps': [64, 32, 16, 8, 4, 2, 1],
    'min_dim': 512,
    'steps': [8, 16, 32, 64, 128, 256, 512],
    'min_sizes': [35.84, 76.8, 153.6, 230.4, 307.2, 384.0, 460.0],
    'max_sizes': [76.8, 153.6, 230.4, 307.2, 384.0, 460.8, 537.6],
    'aspect_ratios': [[2,3], [2, 3], [2, 3], [2, 3],[2, 3], [2], [2]],
    'variance': [0.1, 0.2],
    'clip': True,
    'name': 'VOC',
}

def build_ssd(phase, size=512, num_classes=4):
    if phase != "test" and phase != "train":
        print("ERROR: Phase: " + phase + " not recognized")
        return
    # if size != 300:
    #     print("ERROR: You specified size " + repr(size) + ". However, " +
    #           "currently only SSD300 (size=300) is supported!")
    #     return
    base_, extras_, head_ = multibox(ResNet(BasicBlock, [2,2,2,2], 3),
                                     add_extras(extras[str(size)], 512),
                                     mbox[str(size)], num_classes)
    return SSD(phase, size, base_, extras_, head_, num_classes)

修改head_和extras_,

extras = {
    '300': [256, 'S', 512, 128, 256, 128, 256],
    '512': [256, 'S', 512, 128, 256, 128, 256, 'S', 256, 'S', 256],
}

mbox = {
    '300': [4, 6, 6, 6, 4, 4],  # number of boxes per feature map location
    '512': [4, 6, 6, 6, 6, 4, 4],
}

参考文献：https://blog.csdn.net/github_36923418/article/details/85226306