SketchEmbedNet: Learning Novel Concepts by Imitating Drawings
arXiv:2009.04806
Abstract
Sketch drawings capture the salient information of visual concepts. Previous work has shown that neural networks are capable of producing sketches of natural objects drawn from a small number of classes. While earlier approaches focus on generation quality or retrieval, we explore properties of image representations learned by training a model to produce sketches of images. We show that this generative, class-agnostic model produces informative embeddings of images from novel examples, classes, and even novel datasets in a few-shot setting. Additionally, we find that these learned representations exhibit interesting structure and compositionality.
arXiv commentICML 2021
References in corpus (14)
- Efficient Estimation of Word Representations in Vector Space
- Auto-Encoding Variational Bayes
- Prototypical Networks for Few-shot Learning
- Conditional Image Generation with PixelCNN Decoders
- ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness
- A Neural Representation of Sketch Drawings
- Unsupervised Learning via Meta-Learning
- End-to-End Photo-Sketch Generation via Fully Convolutional Representation Learning
- One-Shot Generalization in Deep Generative Models
- Unsupervised Meta-Learning For Few-Shot Image Classification
- Why Do Line Drawings Work? A Realism Hypothesis
- Sketch-pix2seq: a Model to Generate Sketches of Multiple Categories
- Few-shot Autoregressive Density Estimation: Towards Learning to Learn Distributions
- Assume, Augment and Learn: Unsupervised Few-Shot Meta-Learning via Random Labels and Data Augmentation