ID 原文 译文
1293 为了更好的提升两阶段目标检测的精度与速度性能，提出了一种基于迁移学习方法的融合深度扩张卷积网络和轻量化网络的检测模型。 In order to improve the accuracy and speed performance of two-stage object detection, this paper proposes a detection model based ontransfer learning method that fuses the deep dilated convolutions network and the light-weight network.
1294 首先用扩张卷积网络替换主干网络中部分的卷积残差模块———深度扩张卷积网络 D_dNet-65; First, the dilated con-volutions network is used to replace the convolutional residual module in the backbone network, namely deep dilated convo-lution network(D_dNet-65).
1295 然后对预训练后的特征图进行压缩操作，并增加一个 81 类的全连接层以确保正常进行分类和回归操作———轻量化网络结构; Then, by compressing the pretrained feature map and adding an 81-class fully connected layerto replace the original two layers, namely light-weight network.
1296 最后，引入迁移学习方法并融合 D_dNet 和轻量化网络结构，通过迁移实现模型的进一步优化。 Finally, the transfer learning method is introduced in the pre-training to optimize the model (D_dNet and light-weight network).
1297 实验在典型的数据集 MSCOCO以及 VOC07 上进行。 The experiment was carried out on a typical data set, MSCOCO and VOC07.
1298 实验评估表明，本文提出的方法具有良好的有效性和可扩展性。 And the experiment shows that the method proposed in this paper has good effectiveness and scal-ability.
1299 在低照度环境下采集的图像具有低信噪比、低对比度及低分辨率等特点，导致图像难以识别利用。 The images acquired in the low illumination environment have the characteristics of low signal-to-noise ra-tio, low contrast and low resolution, which make the image difficult to identify and utilize.
1300 为了提升低照度图像的质量，本文提出一种基于 U-Net 生成对抗网络的低照度图像增强方法。 In order to improve the imagequality of low-light images, this paper proposes a low-light image enhancement method based on U-net generative adversarial network (GAN).
1301 首先利用 U-Net 框架实现生成对抗网络中的生成网络，然后利用该生成对抗网络学习从低照度图像到正常照度图像的特征映射，最终实现低照度图像的照度增强。 First, the U-net framework is used to implement the generative network of GAN, and then the GAN is usedto learn the feature mapping from the low-light image to the normal-light image, and ultimately achieve illumination en-hancement for the low-light image. Finally, this method is verified by experiments.
1302 实验结果表明，与主流算法相比，本文提出的方法能够更有效的提升低照度图像的亮度与对比度。 The experimental results show that, com-pared with the mainstream algorithm, the proposed algorithm can effectively improve the brightness and contrast of low-lightimage.