Research map / Detection and segmentation / Open-set and few-shot detection
Variational Amodal Object Completion
Huan Ling, David Acuna, Karsten Kreis, Seung Wook Kim, Sanja Fidler
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 full-mask-only pretraining of decoder (ablation) | Amodal-VAE · KINS test | Invisible mIOU (%): 58.92 -> 62.85 | Improved |
| add simulated partial-to-full training (ablation) | Amodal-VAE · KINS test | Invisible mIOU (%): 35.82 -> 62.85 | Improved |
| add partial-mask-only likelihood finetuning (ablation) | Amodal-VAE · KINS test | Invisible mIOU (%): 57.04 -> 62.85 | Improved |
| add latent code distance loss (ablation) | Amodal-VAE · KINS test | Invisible mIOU (%): 56.90 -> 62.85 | Improved |
| add class conditioning (ablation) | Amodal-VAE · KINS test | Invisible mIOU (%): 53.03 -> 62.85 | Improved |
| add RGB image features (ablation) | Amodal-VAE · KINS test | Invisible mIOU (%): 62.85 -> 61.97 | Hurt |
| sample 20 posterior latents and select by visible-area IoU instead of using posterior mode (ablation) | Amodal-VAE · KINS test | Invisible mIOU (%): 62.85 -> 62.98 | No clear effect |
| sample 20 posterior latents and select best by amodal GT IoU (ablation) | Amodal-VAE · KINS test | Invisible mIOU (%): 62.85 -> 69.96 | Improved |
| provide GT amodal bounding box crop | Amodal-VAE · KINS test | Invisible mIOU (%): 62.85 -> 82.30 | Improved |
| use Amodal-VAE instead of De-occlusion | inpainting pipeline · KINS inpainting | FID: 50.36 -> 41.44 | Improved |
| use Amodal-VAE instead of De-occlusion (trained on Cityscapes) | completion · KINS | Invisible mIOU (%): 48.23 -> 56.18 | Improved |
Closest papers by meaning
- Amodal Scene Analysis via Holistic Occlusion Relation Inference and Generative Mask Completion (AAAI 2024 · similarity 0.89)
- Open-World Amodal Appearance Completion (CVPR 2025 · similarity 0.89)
- Using Diffusion Priors for Video Amodal Segmentation (CVPR 2025 · similarity 0.88)
- Amodal Ground Truth and Completion in the Wild (CVPR 2024 · similarity 0.88)
- Segment Anything, Even Occluded (CVPR 2025 · similarity 0.86)
- Completing Visual Objects via Bridging Generation and Segmentation (ICML 2024 · similarity 0.85)
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.
