Research map / Detection and segmentation / Open-set and few-shot detection
SMILe: Leveraging Submodular Mutual Information For Robust Few-Shot Object Detection
Anay Majee, Ryan X Sharp, Rishabh Iyer
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 |
|---|---|---|---|
| add SMILe objective during few-shot adaptation | FSCE · PASCAL VOC split 3, 10-shot | novel mAP50: 57.7 -> 59.0 | Improved |
| add SMILe objective during few-shot adaptation | DiGeo · PASCAL VOC split 3, 10-shot | novel mAP50: 52.9 -> 56.4 | Improved |
| add SMILe objective | PDC (imTED, ViT-B) · MS-COCO 30-shot | novel mAP50: 47.3 -> 49.9 | Improved |
| add SMILe objective | PDC (imTED, ViT-B) · MS-COCO 10-shot | novel mAP: 23.4 -> 25.8 | Improved |
| add SMILe objective | PDC (imTED, ViT-B) · PASCAL VOC split 3, 1-shot | novel mAP50: 18.9 -> 15.1 | Hurt |
| add intra-class total-information term L_comb^intra to FSCE (ablation) | FSCE · PASCAL VOC split 1, 10-shot | base mAP50: 86.1 -> 89.6 | Improved |
| add inter-class SMI term L_comb^inter to FSCE (ablation) | FSCE · PASCAL VOC split 1, 10-shot | novel mAP50: 57.8 -> 61.0 | Improved |
| add both inter and intra terms vs intra only (ablation) | FSCE · PASCAL VOC split 1, 10-shot | novel mAP50: 60.1 -> 61.1 | Improved |
| use Facility-Location / FLMI instead of Graph-Cut / GCMI (ablation) | DiGeo · PASCAL VOC split 1, 10-shot | novel mAP50: 60.9 -> 62.3 | Improved |
| use Facility-Location / FLMI instead of Graph-Cut / GCMI (ablation) | FSCE · PASCAL VOC split 1, 10-shot | novel mAP50: 60.3 -> 61.1 | Improved |
| add SMILe to AGCM | AGCM · base class retention | base mAP retention: ~3.5% better retention | Improved |
| add SMILe to AGCM | AGCM · confusion matrix | class confusion: 11% lower confusion | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- DiGeo — DiGeo: Discriminative Geometry-Aware Learning for Generalized Few-Shot Object Detection (CVPR 2023) · better on 2 of 2 shared results
- FSCE — FSCE: Few-Shot Object Detection via Contrastive Proposal Encoding (CVPR 2021) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Few-Shot Object Detection via Classification Refinement and Distractor Retreatment (CVPR 2021 · similarity 0.89)
- FSOD-VFM: Few-Shot Object Detection with Vision Foundation Models and Graph Diffusion (ICLR 2026 · similarity 0.88)
- Multi-Scale Positive Sample Refinement for Few-Shot Object Detection (ECCV 2020 · similarity 0.88)
- Label, Verify, Correct: A Simple Few Shot Object Detection Method (CVPR 2022 · similarity 0.88)
- Fine-Grained Prototypes Distillation for Few-Shot Object Detection (AAAI 2024 · similarity 0.87)
- Weakly Supervised Few-Shot Object Detection with DETR (AAAI 2024 · similarity 0.87)
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.
