Caltech-UCSD Birds-200-鸟类数据集

 

数据集包含 200 种鸟类（主要为北美洲鸟类）照片的图像数据集，可用于图像识别工作。分类数量：200；图片数量： 11,788；平均每张图片含有的标注数量：15 个局部位置，312 个二进制属性，1 个边框框。

We present a simple and effective architecture for fine-grained visual recognition called Bilinear Convolutional Neural Networks (B-CNNs). These networks represent an image as a pooled outer product of features derived from two CNNs and capture localized feature interactions in a translationally invariant manner. B-CNNs belong to the class of orderless texture representations but unlike prior work they can be trained in an end-to-end manner. Our most accurate model obtains 84.1%, 79.4%, 86.9% and 91.3% per-image accuracy on the Caltech-UCSD birds, NABirds, FGVC aircraft, and Stanford cars dataset