Diversity Transfer Network for Few-Shot Learning
Authors: Mengting Chen, Yuxin Fang, Xinggang Wang, Heng Luo, Yifeng Geng, Xinyu Zhang, Chang Huang, Wenyu Liu, Bo Wang10559-10566
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We perform extensive experiments and ablation studies on three datasets, i.e., mini Image Net, CIFAR100 and CUB. The results show that DTN, with single-stage training and faster convergence speed, obtains the state-of-the-art results among the feature generation based few-shot learning methods. |
| Researcher Affiliation | Collaboration | 1School of EIC, Huazhong University of Science and Technology 2Horizon Robotics, Inc. 3Vector Institute 4PMCC, UHN |
| Pseudocode | No | The paper does not include any pseudocode blocks or algorithm sections. |
| Open Source Code | Yes | Code and supplementary material are available at: https://github.com/Yuxin-CV/DTN. |
| Open Datasets | Yes | The mini Image Net dataset has been widely used by few-shot learning since it is firstly proposed by Vinyals et al.. The CIFAR100 dataset (Krizhevsky and Hinton 2009) contains 6000 images of 100 classes. The CUB dataset (Wah et al. 2011) is a finegrained dataset from 200 categories of birds. |
| Dataset Splits | Yes | The mini Image Net dataset has been widely used by few-shot learning... There are 64, 16 and 20 classes for training, validation, and testing respectively. The CIFAR100 dataset... We use 64, 16, and 20 classes for training, validation, and testing, respectively. The CUB dataset... It is divided into training, validation, and testing sets with 100, 50, and 50 categories respectively. The splits of CIFAR100 and CUB follow Schwartz et al.. |
| Hardware Specification | No | The paper does not provide specific hardware details such as GPU models, CPU types, or memory amounts used for the experiments. |
| Software Dependencies | No | The paper describes the model architecture and training strategy but does not list specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions). |
| Experiment Setup | Yes | The feature extractor for DTN is a CNN with 4 convolutional modules. Each module contains a 3 3 convolutional layer with 64 channels followed by a batch normalization(BN) layer, a Re LU non-linearity layer, and a 2 2 max-pooling layer... The mapping function φ1... is a fully-connected (FC) layer with 2048 units followed by a leaky Re LU activation (max(x, 0.2x)) layer, and a dropout layer with 30% dropout rate... We choose T = 5 and γ = [0, 0, 1, 1, 2, 2] for training DTN... |