Research map / Vision-language and multimodal / Human and embodied modeling
Human-in-the-Loop Visual Re-ID for Population Size Estimation
Gustavo Perez, Daniel Sheldon, Grant Van Horn, Subhransu Maji
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 |
|---|---|---|---|
| use similarity-driven nested importance sampling instead of uniform nested Monte Carlo (ablation) | population size estimator · 7 animal Re-ID datasets | average error (%): 66.9% -> 12.3% | Improved |
| use human-in-the-loop nested importance sampling instead of k-means with elbow heuristic | Re-ID population size estimation · 7 animal Re-ID datasets | average error (%): 80.4% -> 12.3% | Improved |
| add pairwise constraints to k-means (pck-means) | k-means with elbow · 7 animal Re-ID datasets | average error (%): 80.4% -> 75.4% | No clear effect |
| use farthest-first traversal baseline instead of clustering baselines | population size estimation · 7 animal Re-ID datasets | average error (%): 75.4% -> 38.2% (pck-means vs FFT) | Improved |
| increase sampled-neighbor to sampled-vertex ratio r=M/N from 0.5 to 7 (ablation) | Nested-IS · MacaqueFaces | human effort for near-zero error: close-to-zero error with 1/5 of the sampled pairs | Improved |
| use smaller MegaDescriptor-B-224 instead of L-384 embeddings (ablation) | Nested-IS · MacaqueFaces | estimation error | No clear effect |
| estimate k with Nested-IS before clustering instead of spending effort on active clustering | k-means clustering · Re-ID and fine-grained classification clustering accuracy | clustering accuracy | Improved |
| use estimated confidence intervals from nested importance sampling | Nested-IS · 7 animal Re-ID datasets | CI coverage: close to 95% coverage at about 0.02% of pairs | Improved |
Closest papers by meaning
- Active Learning for Animal Re-Identification with Ambiguity-Aware Sampling (AAAI 2026 · similarity 0.87)
- OpenAnimals: Revisiting Person Re-Identification for Animals Towards Better Generalization (ICCV 2025 · similarity 0.86)
- A Systematic Evaluation and Benchmark for Person Re-Identification: Features, Metrics, and Datasets (TPAMI 2018 · similarity 0.86)
- Large-Scale Training Data Search for Object Re-Identification (CVPR 2023 · similarity 0.85)
- Alice Benchmarks: Connecting Real World Re-Identification with the Synthetic (ICLR 2024 · similarity 0.84)
- Generating the Wild: Individual-Consistent Image-to-Video Generation for Wildlife (NeurIPS 2026 · similarity 0.82)
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.
