Research map / Detection and segmentation / CNN features
MVTec D2S: Densely Segmented Supermarket Dataset
Patrick Follmann, Tobias Bottger, Philipp Hartinger, Rebecca Konig, Markus Ulrich
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 10,000 artificially assembled images to train (ablation) | Mask R-CNN · D2S test (instance segmentation) | AP: 49.5 -> 79.9 | Improved |
| add 10,000 artificially assembled images to train (ablation) | FCIS · D2S test (instance segmentation) | AP: 45.6 -> 72.5 | Improved |
| train on artificial aug images alone instead of train (ablation) | Mask R-CNN · D2S test (instance segmentation) | AP: 49.5 -> 71.6 | Improved |
| train on aug alone instead of train (ablation) | FCIS · D2S test (instance segmentation) | AP: 45.6 -> 69.8 | Improved |
| train on all ten rotations instead of the first rotation (ablation) | Mask R-CNN · D2S test (instance segmentation) | AP: 33.8 -> 49.5 | Improved |
| train on all ten rotations instead of the first rotation (ablation) | FCIS · D2S test (instance segmentation) | AP: 36.5 -> 45.6 | Improved |
| train on all three lightings instead of default lighting (ablation) | Mask R-CNN · D2S test (instance segmentation) | AP: 46.1 -> 49.5 | Improved |
| train on all three lightings instead of default lighting (ablation) | FCIS · D2S test (instance segmentation) | AP: 46.0 -> 45.6 | No clear effect |
| train on both rotation and lighting variation (ablation) | Mask R-CNN · D2S test (instance segmentation) | AP: 36.3 -> 49.5 | Improved |
| train on both rotation and lighting variation (ablation) | FCIS · D2S test (instance segmentation) | AP: 32.7 -> 45.6 | Improved |
| add artificial images to train (ablation) | FCIS (box AP) · D2S test (object detection, box AP) | AP: 69.9 -> 68.3 | Hurt |
| add artificial images to train (ablation) | RetinaNet (box AP) · D2S test (object detection, box AP) | AP: 51.0 -> 80.1 | Improved |
Closest papers by meaning
- OVFoodSeg: Elevating Open-Vocabulary Food Image Segmentation via Image-Informed Textual Representation (CVPR 2024 · similarity 0.85)
- ScaleNet: Guiding Object Proposal Generation in Supermarkets and Beyond (ICCV 2017 · similarity 0.84)
- MUVA: A New Large-Scale Benchmark for Multi-View Amodal Instance Segmentation in the Shopping Scenario (ICCV 2023 · similarity 0.84)
- P2Seg: Pointly-supervised Segmentation via Mutual Distillation (ICLR 2024 · similarity 0.84)
- Scaling up Image Segmentation across Data and Tasks (CVPR 2025 · similarity 0.83)
- SANSA: Unleashing the Hidden Semantics in SAM2 for Few-Shot Segmentation (NeurIPS 2025 · similarity 0.83)
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.
