Research map / Architectures and efficiency / Neural architecture search
Compatibility-Aware Heterogeneous Visual Search
Rahul Duggal, Hao Zhou, Shuo Yang, Yuanjun Xiong, Wei Xia, Zhuowen Tu, Stefano Soatto
Check 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 the homogeneous ResNet-101 query with Cmp-NAS while retaining ResNet-101 for gallery indexing | Visual search · IJB-C face retrieval | TPIR@FPIR=10^-1 (%): 86.7 -> 85.1 | Hurt |
| Replace the homogeneous MobileNetV2 system with a Cmp-NAS query and ResNet-101 gallery | Visual search · IJB-C face retrieval | TPIR@FPIR=10^-1 (%): 77.1 -> 85.1 | Improved |
| Replace the homogeneous ResNet-101 query with Cmp-NAS while retaining ResNet-101 for gallery indexing | Visual search · DeepFashion2 retrieval | Top-10 retrieval accuracy (%): 65.2 -> 64.9 | Hurt |
| Replace the homogeneous MobileNetV3 system with a Cmp-NAS query and ResNet-101 gallery | Visual search · DeepFashion2 retrieval | Top-10 retrieval accuracy (%): 62.7 -> 64.9 | Improved |
| Add BCT when training the query model (ablation) | Pruned ResNet-101 query model · IJB-C face retrieval | Heterogeneous 1:N accuracy: Magnitude prune: 0.0 -> 87.2; Channel prune: 0.0 -> 87.4 | Improved |
| Replace training from scratch with fine-tuning (ablation) | Magnitude-pruned ResNet-101 query · IJB-C face retrieval | Heterogeneous 1:N accuracy: 0.0 -> 86.5 | Improved |
| Replace training from scratch with knowledge distillation (ablation) | Magnitude-pruned ResNet-101 query · IJB-C face retrieval | Heterogeneous 1:N accuracy: 0.0 -> 0.0 | No clear effect |
| Replace BCT with fine-tuning (ablation) | Magnitude-pruned ResNet-101 query · IJB-C face retrieval | Heterogeneous 1:N accuracy: 87.2 -> 86.5 | Hurt |
| Replace BCT with knowledge distillation (ablation) | Magnitude-pruned ResNet-101 query · IJB-C face retrieval | Heterogeneous 1:N accuracy: 87.2 -> 0.0 | Hurt |
| Replace training from scratch with fine-tuning (ablation) | Channel-pruned ResNet-101 query · IJB-C face retrieval | Heterogeneous 1:N accuracy: 0.0 -> 86.3 | Improved |
| Replace training from scratch with knowledge distillation (ablation) | Channel-pruned ResNet-101 query · IJB-C face retrieval | Heterogeneous 1:N accuracy: 0.0 -> 0.0 | No clear effect |
| Replace BCT with fine-tuning (ablation) | Channel-pruned ResNet-101 query · IJB-C face retrieval | Heterogeneous 1:N accuracy: 87.4 -> 86.3 | Hurt |
Closest papers by meaning
- Learning Compatible Embeddings (ICCV 2021 · similarity 0.88)
- Exploiting Distribution Constraints for Scalable and Efficient Image Retrieval (ICLR 2025 · similarity 0.86)
- Aligning Vision Models with Human Aesthetics in Retrieval: Benchmarks and Algorithms (NeurIPS 2024 · similarity 0.85)
- NUDGE: Lightweight Non-Parametric Fine-Tuning of Embeddings for Retrieval (ICLR 2025 · similarity 0.85)
- Stationary Representations: Optimally Approximating Compatibility and Implications for Improved Model Replacements (CVPR 2024 · similarity 0.84)
- Learning to Align Generative Appearance Priors for Fine-grained Image Retrieval (NeurIPS 2026 · similarity 0.84)
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.
