DenseCap：一种对于密集抓图的全卷积定位神经网络

         

      以下是对这边paper的翻译，仅当练手

 DenseCap: Fully Convolutional Localization Networks for Dense Captioning

We introduce the dense captioning task, which requires a computer vision system to both localize and describe salient regions in images in natural language. The dense captioning task generalizes object detection when the descriptions consist of a single word, and Image Captioning when one predicted region covers the full image. To address the localizationand description task jointly we propose a Fully ConvolutionalLocalization Network (FCLN) architecture that processes an image with a single, efficient forward pass, requires no external regions proposals, and can be trained end-to-end with a single round of optimization. The architectureis composed of a Convolutional Network, a noveldense localization layer, and Recurrent Neural Network language model that generates the label sequences. We evaluate our network on the Visual Genome dataset, which comprises 94,000 images and 4,100,000 region-grounded captions. We observe both speed and accuracy improvements over baselines based on current state of the art approachesin both generation and retrieval settings.

    我介绍一种密集侦测的模型，这个要求计算机视觉系统可以定位和用自然语言描述图中辨识到的区域。密集侦测技术侦测到对象并用简单的单次进行描述，并标注预测的区域。我们提出了一种全卷积定位神经网络，简称FCLN ,这个结构可以单一的，有效率向前传播数据来处理图像，不需要额外的区域参考，并且端对端的被训练。该结构由卷积网络，一个特别的定位层，和一个可以产生标签序列的循环网络语言模型。我们使用 Visual Genome dataset 评估了我们的网络，这个数据集包括94000张图片和410000 region-grounded captions. （看了数据结构才能翻译）。我们监控了速度和精度