SketchEmbedNet: Learning Novel Concepts by Imitating Drawings
Authors: Alexander Wang, Mengye Ren, Richard Zemel
ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we present our experiments on Sketch Embed Net and investigate the properties of Sketch Embeddings. We study the generalization of Sketch Embeddings through classification tasks involving novel examples, classes and datasets. |
| Researcher Affiliation | Academia | 1University of Toronto, Toronto, Canada 2Vector Institute 3CIFAR. Correspondence to: Alexander Wang <alexw@cs.toronto.edu>, Mengye Ren <mren@cs.toronto.edu>, Richard S. Zemel <zemel@cs.toronto.edu>. |
| Pseudocode | No | No structured pseudocode or algorithm blocks were found in the paper. |
| Open Source Code | No | The paper does not provide any explicit statements about open-source code release, nor does it include links to a code repository. |
| Open Datasets | Yes | Quickdraw (Jongejan et al., 2016) (Figure 2a) pairs sketches with a line drawing rendering of the motor program and contains 345 classes of 70,000 examples, produced by human players participating in the game Quick, Draw! ... Sketchy (Sangkloy et al., 2016) (Figure 2b) is a more challenging collection of (photorealistic) natural image sketch pairs and contains 125 classes from Image Net (Deng et al., 2009)... |
| Dataset Splits | Yes | Data processing procedures and class splits are in Appendix C. ... Following the few-shot learning protocol of Vinyals et al. (2016) and Ravi & Larochelle (2017), datasets are split into disjoint sets of training, validation, and test classes. For the Quickdraw dataset, we use 300 classes for training, and 20 classes for validation and 25 for testing. |
| Hardware Specification | No | The paper mentions training models but does not provide specific hardware details such as GPU or CPU models, memory specifications, or cloud computing instance types. |
| Software Dependencies | No | The paper mentions using the Adam optimizer but does not specify version numbers for any software dependencies, libraries, or programming languages used in the experiments. |
| Experiment Setup | Yes | Sketch Embed Net is trained for 300k iterations with batch size of 256 for Quickdraw and 64 for Sketchy due to memory constraints. Initial learning rate is 1e-3 decaying by 0.85 every 15k steps. We use the Adam (Kingma & Ba, 2015) optimizer and clip gradient values to 1.0. Latent space dim(z) = 256, RNN output size is 1024, and hypernetwork embedding is 64. Mixture count is M = 30 and Gaussian blur from Lpixel uses σ = 2.0. Conv4 encoder is identical to Vinyals et al. (2016) and the Res Net-12 encoder uses 4 blocks of 64-128-256-512 filters with Re LU activations. α is set to 0 and increases by 0.05 every 10k training steps with an empirically obtained cap at αmax = 0.50 for Quickdraw and αmax = 0.75 for Sketchy. See Appendix B for additional details. |