Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,63 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: mit
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
library_name: pytorch
|
| 4 |
+
tags:
|
| 5 |
+
- image-classification
|
| 6 |
+
- data-augmentation
|
| 7 |
+
- small-sample-learning
|
| 8 |
+
datasets:
|
| 9 |
+
- cifar100
|
| 10 |
+
- cifar10
|
| 11 |
+
metrics:
|
| 12 |
+
- accuracy
|
| 13 |
+
---
|
| 14 |
+
|
| 15 |
+
# Prior-Guided Augmentation: Experiment Results
|
| 16 |
+
# 先验引导增强：实验结果备份库
|
| 17 |
+
|
| 18 |
+
Official checkpoint and experimental results repository for the paper:
|
| 19 |
+
**"Prior-Guided Augmentation: A Reliable Strategy for Small-Sample Datasets"**
|
| 20 |
+
|
| 21 |
+
This repository contains the official model weights, search logs, and visualizations for the Prior-Guided Augmentation (PGA) framework.
|
| 22 |
+
本仓库包含该论文配套的官方模型权重、搜索日志以及可视化图表。
|
| 23 |
+
|
| 24 |
+
---
|
| 25 |
+
|
| 26 |
+
## 📊 Core Results / 核心实验结果
|
| 27 |
+
|
| 28 |
+
### CIFAR-100 (100-shot)
|
| 29 |
+
| Metric / 指标 | Result / 结果 |
|
| 30 |
+
| :--- | :--- |
|
| 31 |
+
| **Top-1 Accuracy** | 40.74% |
|
| 32 |
+
| **Stability (Std)** | 0.78 (Highest in benchmarks) |
|
| 33 |
+
| **Optimal Policy** | ColorJitter (m=9, p=0.5) |
|
| 34 |
+
|
| 35 |
+
### CIFAR-10 (50-shot)
|
| 36 |
+
| Metric / 指标 | Result / 结果 |
|
| 37 |
+
| :--- | :--- |
|
| 38 |
+
| **Top-1 Accuracy** | 50.0% |
|
| 39 |
+
| **Stability (Std)** | 0.0 (Zero Variance / 零方差) |
|
| 40 |
+
|
| 41 |
+
---
|
| 42 |
+
|
| 43 |
+
## 📂 Repository Structure / 仓库结构
|
| 44 |
+
|
| 45 |
+
- `checkpoints/`: Model weights (`best_model.pth`). / 官方模型权重。
|
| 46 |
+
- `figures/`: Visualizations (Heatmaps, Efficiency plots). / 论文可视化图表（热图、效率对比图等）。
|
| 47 |
+
- [phase_c_final_policy.json](cci:7://file:///Users/qinfuyao/Code/prior-guided-aug/outputs/phase_c_final_policy.json:0:0-0:0): The discovered optimal augmentation policy. / 搜索出的最优增强策略。
|
| 48 |
+
- [cifar10_50shot_results.csv](cci:7://file:///Users/qinfuyao/Code/prior-guided-aug/outputs/cifar10_50shot_results.csv:0:0-0:0): CIFAR-10 experimental data. / CIFAR-10 50-shot 实验原始数据。
|
| 49 |
+
- [stability_seeds_results.csv](cci:7://file:///Users/qinfuyao/Code/prior-guided-aug/outputs/stability_seeds_results.csv:0:0-0:0): Raw data for stability tests across random seeds. / 跨随机种子的稳定性原始测试数据。
|
| 50 |
+
- [destructiveness_metrics.csv](cci:7://file:///Users/qinfuyao/Code/prior-guided-aug/outputs/destructiveness_metrics.csv:0:0-0:0): LPIPS/SSIM analysis for semantic preservation. / 语义保真度分析（LPIPS/SSIM）。
|
| 51 |
+
|
| 52 |
+
---
|
| 53 |
+
|
| 54 |
+
## 📜 Citation / 引用
|
| 55 |
+
If you find this repository useful, please cite our work:
|
| 56 |
+
如果您觉得本工作对您有帮助，请引用我们的工作：
|
| 57 |
+
|
| 58 |
+
(Stay tuned! Citation will be updated upon acceptance. / 论文接收后将在此更新引用信息。)
|
| 59 |
+
|
| 60 |
+
---
|
| 61 |
+
|
| 62 |
+
## 🔗 Links / 相关链接
|
| 63 |
+
- **GitHub**: [imnotnoahhh/prior-guided-aug-policy](https://github.com/imnotnoahhh/prior-guided-aug-policy)
|