Field-Guide-Inspired Zero-Shot Learning
Field-Guide-Inspired Zero-Shot Learning
复制标题
DOI：
10.1109/iccv48922.2021.00941
复制
发表时间：
2021-08
期刊：
影响因子：
--
通讯作者：
Utkarsh Mall;Bharath Hariharan;Kavita Bala
中科院分区：
文献类型：
--
作者：
Utkarsh Mall;Bharath Hariharan;Kavita Bala
现代识别系统需要大量的监督才能实现准确性。适应新的领域需要来自专家的大量数据，这是繁重的，并且可能变得过于昂贵。零触发学习需要一个新类别的注释属性集。在部署过程中，为一个新的类别标注完整的属性集被证明是一项繁琐而昂贵的任务。当识别领域是专家领域时尤其如此。我们引入了一种新的现场指南启发的方法来零杆注释，其中学习者模型交互式地要求定义类的最有用的属性。我们评估了我们的方法与属性注释，如CUB，SUN和AWA2分类基准，并表明我们的模型实现了一个模型的性能与完整的注释在成本显着较少的注释数量。由于专家的时间非常宝贵，因此降低注释成本对于实际部署非常有价值。
Modern recognition systems require large amounts of supervision to achieve accuracy. Adapting to new domains requires significant data from experts, which is onerous and can become too expensive. Zero-shot learning requires an annotated set of attributes for a novel category. Annotating the full set of attributes for a novel category proves to be a tedious and expensive task in deployment. This is especially the case when the recognition domain is an expert domain. We introduce a new field-guide-inspired approach to zero-shot annotation where the learner model interactively asks for the most useful attributes that define a class. We evaluate our method on classification benchmarks with attribute annotations like CUB, SUN, and AWA2 and show that our model achieves the performance of a model with full annotations at the cost of significantly fewer number of annotations. Since the time of experts is precious, decreasing annotation cost can be very valuable for real-world deployment.