Research map / Detection and segmentation / Open-set and few-shot detection
GENESIS-V2: Inferring Unordered Object Representations without Iterative Refinement
Martin Engelcke, Oiwi Parker Jones, Ingmar Posner
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 IC-SBP to infer unordered object representations without recurrent inference | GENESIS-V2 · ObjectsRoom | ARI-FG: 0.63$\pm$0.03 -> 0.85$\pm$0.01 | Improved |
| Use IC-SBP to infer unordered object representations without recurrent inference | GENESIS-V2 · ShapeStacks | ARI-FG: 0.70$\pm$0.05 -> 0.81$\pm$0.01 | Improved |
| Use GENESIS-V2 rather than the generative baselines | GENESIS-V2 · ShapeStacks | MSC-FG: 0.67$\pm$0.01 vs 0.70$\pm$0.05 | Mixed |
| Use GENESIS-V2 rather than GENESIS | GENESIS-V2 · ObjectsRoom | ELBO: -7023$\pm$2 -> -7040$\pm$2 | Hurt |
| Use GENESIS-V2 rather than GENESIS | GENESIS-V2 · ShapeStacks | ELBO: -7082$\pm$15 -> -7019$\pm$2 | Improved |
| Use GENESIS-V2 rather than GENESIS | GENESIS-V2 · APC | ARI: 0.04 -> 0.55 | Improved |
| Use GENESIS-V2 rather than GENESIS | GENESIS-V2 · APC | FID: 183.2 -> 245.6 | Hurt |
| Add an autoregressive prior (ablation) | GENESIS-V2 · ObjectsRoom | ARI-FG: 0.79$\pm$0.01 -> 0.84$\pm$0.01 | Improved |
| Add an autoregressive prior (ablation) | GENESIS-V2 · ShapeStacks | ARI-FG: 0.79$\pm$0.01 -> 0.81$\pm$0.00 | Improved |
| Use semi-convolutional rather than standard convolutional embeddings (ablation) | GENESIS-V2 · ObjectsRoom | ARI-FG: 0.79$\pm$0.05 -> 0.84$\pm$0.01 | Improved |
| Use semi-convolutional rather than standard convolutional embeddings (ablation) | GENESIS-V2 · ShapeStacks | ARI-FG: 0.60$\pm$0.38 -> 0.81$\pm$0.00 | Improved |
| Replace the Gaussian kernel with the Laplacian kernel (ablation) | GENESIS-V2 without autoregressive prior · ObjectsRoom | ARI-FG: 0.79$\pm$0.01 -> 0.74$\pm$0.08 | Hurt |
Later papers that compared with this method
3 later accepted papers report Genesis-v2 as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Latent Noise Segmentation: How Neural Noise Leads to the Emergence of Segmentation and Grouping (ICML 2024 · better on 1 of 1 shared result)
- Slot-VAE: Object-Centric Scene Generation with Slot Attention (ICML 2023 · better on 4 of 6 shared results)
- Multi-Object Representation Learning via Feature Connectivity and Object-Centric Regularization (NeurIPS 2023 · better on 1 of 1 shared result)
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
- GENESIS: Generative Scene Inference and Sampling with Object-Centric Latent Representations (ICLR 2020 · similarity 0.90)
- Unsupervised Part Discovery from Contrastive Reconstruction (NeurIPS 2021 · similarity 0.87)
- gen2seg: Generative Models Enable Generalizable Instance Segmentation (ICLR 2026 · similarity 0.87)
- Conditional Object-Centric Learning from Video (ICLR 2022 · similarity 0.87)
- Using Diffusion Priors for Video Amodal Segmentation (CVPR 2025 · similarity 0.86)
- Hierarchy-Agnostic Unsupervised Segmentation: Parsing Semantic Image Structure (NeurIPS 2024 · similarity 0.86)
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.
