Research map / Adversarial robustness and security / Attacks on text and graphs
An Adversarial Approach to Hard Triplet Generation
Yiru Zhao, Zhongming Jin, Guo-jun Qi, Hongtao Lu, Xian-sheng Hua
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace OHEM mining with adversarial hard triplet generation (HTG) | ResNet-18 triplet+softmax · CARS196 | Recall@1 (%): 68.2 -> 70.6 | Improved |
| add OHEM hard mining over random sampling | ResNet-18 triplet+softmax · CARS196 | Recall@1 (%): 66.6 -> 68.2 | Improved |
| add HTG over random sampling | ResNet-18 triplet+softmax · CARS196 | Recall@1 (%): 66.6 -> 70.6 | Improved |
| remove the bias from the softmax layer (ablation) | ResNet-18 with HTG · CARS196 | Recall@1 (%): 69.3 -> 70.6 | Improved |
| remove the bias from the softmax layer (ablation) | ResNet-18 with random sampling · CARS196 | Recall@1 (%): 65.4 -> 66.6 | Improved |
| remove the bias from the softmax layer (ablation) | ResNet-18 with OHEM · CARS196 | Recall@1 (%): 67.1 -> 68.2 | Improved |
| add keypoint maps (KPM) local branch to HTG (ablation) | ResNet-18 with HTG · CARS196 | Recall@1 (%): 70.6 -> 76.5 | Improved |
| add K-Branch local branch to HTG (ablation) | ResNet-18 with HTG · CARS196 | Recall@1 (%): 70.6 -> 72.7 | Improved |
| add ResAttention to HTG (ablation) | ResNet-18 with HTG · CARS196 | Recall@1 (%): 70.6 -> 73.9 | Improved |
| add OHEM mining on attention models (vs random) | KPM model · CARS196 | Recall@1 (%): 72.7 -> 72.5 | No clear effect |
| add OHEM mining on attention models (vs random) | K-Branch model · CARS196 | Recall@1 (%): 71.9 -> 72.2 | No clear effect |
| HTG versus HDC | HTG · CUB200 | Recall@1 (%): 53.6 -> 59.5 | Improved |
Closest papers by meaning
- Triple Generative Adversarial Nets (NeurIPS 2017 · similarity 0.88)
- Triple Generative Adversarial Networks (TPAMI 2021 · similarity 0.87)
- Adversarial Examples Generation for Deep Product Quantization Networks on Image Retrieval (TPAMI 2022 · similarity 0.86)
- Hard negative examples are hard, but useful (ECCV 2020 · similarity 0.86)
- Mathematical Justification of Hard Negative Mining via Isometric Approximation Theorem (ICLR 2024 · similarity 0.85)
- Difficulty Controlled Diffusion Model for Synthesizing Effective Training Data (AAAI 2026 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
