OSM: An open set matting framework with OOD detection and few-shot learning
Bibliographic record
Abstract
Natural image matting is the task of precisely estimating alpha mattes to separate foreground objects from background images.Existing matting methods only focus on classical closed-set problems where object categories and data distributions are similar between training and test sets.However, in the open world setup, there exists a situation where testing samples are drawn from a different distribution than the training data.To handle this situation, we present the first open set matting (OSM) framework that contains two networks: (1) an out-of-distribution (OOD) detection network to identify OOD to-be-matted objects; and (2) an incremental few-shot learning matting module to enlarge the existing knowledge base of to-be-matted objects.Our OOD detection network leverages metric-based prototype learning to be aware of unseen objects and increase inter-class separability, utilizing intra-batch connections to enhance intra-class compactness.Compared to other OOD detection methods, our network achieves state-of-the-art performance on SIMD dataset.Further, our incremental few-shot learning matting module improves the performance on unseen to-be-matted objects by gradually incorporating novel classes into the existing knowledge base without catastrophic forgetting and overfitting.i vi 6 Conclusions 40 7 Appendix 41 vii List of Tables 4.1 OOD detection results on SIMD dataset. . . . . . . . . . . . . . . . . . . . . .4.2 Ablation study results of our OOD detection network on SIMD dataset.PL refers to prototype learning. . . . . . . . . . . . . . . . . . . . . . . . . . . . .4.3 Matting results on SIMD dataset. . . . . . . . . . . . . . . . . . . . . . . . . .4.4 Detailed quantitive matting results of 20 classes of SIMD dataset on SAD metric.Bolden classes are OOD classes, otherwise classes are ID classes. . .4.5 Ablation study results of our incremental few-shot learning matting module on SIMD dataset.Note that Reg is the regularization term based on Elastic Weight Consolidation (EWC). . . . . . . . . . . . . . . . . . . . . . . .7.1 Additional OOD detection results on SIMD dataset. . . . . . . . . . . . . . .
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.005 | 0.003 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.005 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".