Awesome
BOIL: Towards Representation Change for Few-shot Learning
This repository is the official implementation of "BOIL: Towards Representation Change for Few-shot Learning" Our implementations are relied on Torchmeta.
Requirements
We run our code in the following environment using Anaconda.
- Python >= 3.5
- Pytorch == 1.4
- torchvision == 0.5
If you use Pytorch version above 1.5 (which is the latest version at this moment) and torchvision above 0.6, you may encounter problem. In that case, you are encouraged to change to the version in our environment.
To install requirements:
pip install -r requirements.txt
Training
If you want to train 4conv network in the paper, run this command:
./run_4conv.sh
If you want to train ResNet-12 in the paper, run this command:
./run_resnet.sh
If you want to see and change the arguments of training code, run this command:
python3 main.py --help
Evaluation
To evaluate the model(s) and see the results, please refer to the analysis.ipynb
Results
All results were reproduced by our group and reported as the average and standard deviation of the accuracies over 5x1000 tasks.
The values in parenthesis are the number of shots.
1. 5-Way k-shot test accuracy (%) of 4conv network on various benchmark dataset.
<table align="center" style="margin: 0px auto;"> <tr> <td>Domain</td> <td colspan="2">General Domain</td> <td colspan="2">Specific Domain</td> </tr> <tr> <td>Dataset</td> <td>miniImageNet</td> <td>tieredImageNet</td> <td>CUB</td> <td>Cars</td> </tr> <tr> <td>MAML(1)</td> <td><img src="https://latex.codecogs.com/gif.latex?48.47 \pm 0.26" /></td> <td><img src="https://latex.codecogs.com/gif.latex?48.80 \pm 0.34" /></td> <td><img src="https://latex.codecogs.com/gif.latex?53.70 \pm 0.42" /></td> <td><img src="https://latex.codecogs.com/gif.latex?38.16 \pm 0.20" /></td> </tr> <tr> <td>BOIL(1)</td> <td><img src="https://latex.codecogs.com/gif.latex?49.65 \pm 0.19" /></td> <td><img src="https://latex.codecogs.com/gif.latex?50.00 \pm 0.35" /></td> <td><img src="https://latex.codecogs.com/gif.latex?60.45 \pm 0.45" /></td> <td><img src="https://latex.codecogs.com/gif.latex?65.11 \pm 0.36" /></td> </tr> <tr> <td>MAML(5)</td> <td><img src="https://latex.codecogs.com/gif.latex?60.36 \pm 0.25" /></td> <td><img src="https://latex.codecogs.com/gif.latex?64.27 \pm 0.27" /></td> <td><img src="https://latex.codecogs.com/gif.latex?65.11 \pm 0.07" /></td> <td><img src="https://latex.codecogs.com/gif.latex?45.36 \pm 0.23" /></td> </tr> <tr> <td>BOIL(5)</td> <td><img src="https://latex.codecogs.com/gif.latex?65.32 \pm 0.34" /></td> <td><img src="https://latex.codecogs.com/gif.latex?69.64 \pm 0.20" /></td> <td><img src="https://latex.codecogs.com/gif.latex?74.12 \pm 0.24" /></td> <td><img src="https://latex.codecogs.com/gif.latex?65.70 \pm 0.17" /></td> </tr> </table>2. 5-Way k-shot test accuracy (%) of 4conv network on cross-domain adaptation.
<table align="center" style="margin: 0px auto;"> <tr> <td>Adaptation</td> <td colspan="2">General to general</td> <td colspan="2">General to Specific</td> </tr> <tr> <td>Meta-train</td> <td>tieredImageNet</td> <td>miniImageNet</td> <td>miniImageNet</td> <td>miniImageNet</td> </tr> <tr> <td>Meta-test</td> <td>miniImageNet</td> <td>tieredImageNet</td> <td>CUB</td> <td>Cars</td> </tr> <tr> <td>MAML(1)</td> <td><img src="https://latex.codecogs.com/gif.latex?49.45 \pm 0.31" /></td> <td><img src="https://latex.codecogs.com/gif.latex?52.31 \pm 0.33" /></td> <td><img src="https://latex.codecogs.com/gif.latex?40.36 \pm 0.12" /></td> <td><img src="https://latex.codecogs.com/gif.latex?35.27 \pm 0.11" /></td> </tr> <tr> <td>BOIL(1)</td> <td><img src="https://latex.codecogs.com/gif.latex?51.35 \pm 0.18" /></td> <td><img src="https://latex.codecogs.com/gif.latex?54.09 \pm 0.41" /></td> <td><img src="https://latex.codecogs.com/gif.latex?44.38 \pm 0.11" /></td> <td><img src="https://latex.codecogs.com/gif.latex?37.16 \pm 0.35" /></td> </tr> <tr> <td>MAML(5)</td> <td><img src="https://latex.codecogs.com/gif.latex?65.31 \pm 0.12" /></td> <td><img src="https://latex.codecogs.com/gif.latex?64.88 \pm 0.28" /></td> <td><img src="https://latex.codecogs.com/gif.latex?51.34 \pm 0.24" /></td> <td><img src="https://latex.codecogs.com/gif.latex?44.29 \pm 0.28" /></td> </tr> <tr> <td>BOIL(5)</td> <td><img src="https://latex.codecogs.com/gif.latex?70.76 \pm 0.14" /></td> <td><img src="https://latex.codecogs.com/gif.latex?68.97 \pm 0.24" /></td> <td><img src="https://latex.codecogs.com/gif.latex?60.11 \pm 0.32" /></td> <td><img src="https://latex.codecogs.com/gif.latex?50.92 \pm 0.22" /></td> </tr> </table><table align="center" style="margin: 0px auto;"> <tr> <td>Adaptation</td> <td colspan="2">Specific to general</td> <td colspan="2">Specific to Specific</td> </tr> <tr> <td>Meta-train</td> <td>CUB</td> <td>CUB</td> <td>Cars</td> <td>CUB</td> </tr> <tr> <td>Meta-test</td> <td>miniImageNet</td> <td>tieredImageNet</td> <td>CUB</td> <td>Cars</td> </tr> <tr> <td>MAML(1)</td> <td><img src="https://latex.codecogs.com/gif.latex?31.11 \pm 0.21" /></td> <td><img src="https://latex.codecogs.com/gif.latex?34.14 \pm 0.29" /></td> <td><img src="https://latex.codecogs.com/gif.latex?26.27 \pm 0.10" /></td> <td><img src="https://latex.codecogs.com/gif.latex?31.08 \pm 0.18" /></td> </tr> <tr> <td>BOIL(1)</td> <td><img src="https://latex.codecogs.com/gif.latex?35.11 \pm 0.27" /></td> <td><img src="https://latex.codecogs.com/gif.latex?37.88 \pm 0.23" /></td> <td><img src="https://latex.codecogs.com/gif.latex?33.13 \pm 0.29" /></td> <td><img src="https://latex.codecogs.com/gif.latex?34.51 \pm 0.13" /></td> </tr> <tr> <td>MAML(5)</td> <td><img src="https://latex.codecogs.com/gif.latex?38.74 \pm 0.17" /></td> <td><img src="https://latex.codecogs.com/gif.latex?42.11 \pm 0.23" /></td> <td><img src="https://latex.codecogs.com/gif.latex?30.50 \pm 0.21" /></td> <td><img src="https://latex.codecogs.com/gif.latex?39.74 \pm 0.19" /></td> </tr> <tr> <td>BOIL(5)</td> <td><img src="https://latex.codecogs.com/gif.latex?47.63 \pm 0.29" /></td> <td><img src="https://latex.codecogs.com/gif.latex?49.96 \pm 0.10" /></td> <td><img src="https://latex.codecogs.com/gif.latex?42.52 \pm 0.12" /></td> <td><img src="https://latex.codecogs.com/gif.latex?43.73 \pm 0.23" /></td> </tr> </table>