Learning to Generate Chairs, Tables and Cars with Convolutional Networks

By Alexey Dosovitskiy, Jost Tobias Springenberg, Maxim Tatarchenko, Thomas Brox. 2014

267 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Generative Models

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Generative Models

Citations in the corpus

Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks (2015, background): “A recurrent network approach (Gregor et al., 2015) and a deconvolution network approach (Dosovitskiy et al., 2014) have also recently had some success with generating natural images.”
Adversarial Examples for Generative Models (2017, background): “2 Deep generative models, such as [5], learn to generate a variety of outputs, including handwritten digits, faces [6], realistic scenes [7], videos [8], 3D objects [9], and audio [10].”
Geometric Deep Learning: Going beyond Euclidean data (2016, background): “There have been several recent attempts to try to learn a generative model allowing to synthesize new images [120] and speech waveforms [121].”
SampleRNN: An Unconditional End-to-End Neural Audio Generation Model (2016, methodology): “ adding zeros and then applying a linear convolution. This is sometimes called “perforated” upsampling in the context of convolutional neural networks (CNNs). It was ﬁrst demonstrated to work well in Dosovitskiy et al. (2016) and is a fairly common upsampling technique. 2.2 SAMPLE-LEVEL MODULE The l”
Pix3D: Dataset and Methods for Single-Image 3D Shape Modeling (2018, background): “Chairs cover the significant part of Pix3D, because they are common, highly diverse, and well-studied by recent literature [11, 59, 20].”
Octree Generating Networks: Efficient Convolutional Architectures for High-resolution 3D Outputs (2017, neutral): “Similar to the two-dimensional case [9], the outputs are accurate in the overall shape, but lack some fine details.”
Temporal Generative Adversarial Nets with Singular Value Clipping (2016, background): “[3] have proposed models that respectively use recurrent and deconvolutional networks, and successfully generated natural images.”
Synthesizing 3D Shapes via Modeling Multi-view Depth Maps and Silhouettes with Deep Generative Networks (2017, background): “[5] explored generating shapes in multiple viewpoints, though they did not look into 3D shape synthesis, either.”
Variational Autoencoder for Deep Learning of Images, Labels and Captions (2016, methodology): “Our DGDN-CNN model for images is related to but distinct from prior convolutional variational auto-encoder networks [13, 14, 15].”
Semi-Supervised Learning with Context-Conditional Generative Adversarial Networks (2016, methodology): “Dosovitskiy et al. (2015) trained a CNN to generate objects with different shapes, viewpoints and color.”
Baby steps towards few-shot learning with multiple semantics (2019, unknown): “…domain transfer by borrowing examples from relevant categories [28] or using semantic vocabularies [3, 14]; (iv) rendering synthetic examples [9, 35, 48]; (v) augmenting the training examples using geometric and photometric transformations [23] or learning adaptive augmentation strategies…”
DeepVoxels: Learning Persistent 3D Feature Embeddings (2018, background): “There also exists a large body of work on learning lowdimensional embeddings of images that can be decoded to novel views [54, 61, 7, 9, 60, 45].”
Generative Adversarial Active Learning (2017, background): “Conditional GAN(Gauthier, 2014; Dosovitskiy et al., 2014; Mirza and Osindero, 2014) is another variant of GAN in which the generator and discriminator can be conditioned on other variables, e.g., the labels of images.”