Research map / 3D vision / 3D generation
Rosetum3D: A Large-Scale 3D Vision Dataset from Preharvest Roses
Songyan Liu, Xipei Liu, Yujie Liu, Jiali Wu, Xiaofei Liu
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 |
|---|---|---|---|
| fine-tune depth models on Rosetum3D instead of zero-shot NYU-trained (ablation) | monocular depth models · Rosetum3D monocular depth | AbsRel: 0.498 -> 0.165 (best zero-shot Depth Anything vs best fine-tuned BinsFormer) | Improved |
| use ground-truth boxes instead of YOLOv5 detections for top-down keypoints (ablation) | ViTPose · Rosetum3D 2D keypoint localization | AP: 21.9 -> 42.1 | Improved |
| top-down vs bottom-up keypoint methods with detected boxes | ViTPose vs DEKR+HRNet · Rosetum3D 2D keypoint localization | AP: 24.9 -> 21.9 (top-down) | Hurt |
| add multi-frame temporal optimization to 3D localization (ablation) | ViTPose + DETR · Rosetum3D 3D rose localization | MPJPE (mm): 556.8 -> 539.1 | Improved |
| add multi-frame temporal optimization to 3D localization (ablation) | RTMPose + DETR · Rosetum3D 3D rose localization | MPJPE (mm): 558.4 -> 542.2 | Improved |
| use predicted depth instead of ground-truth depth for 3D keypoints | ViTPose + DETR · Rosetum3D 3D rose localization | MPJPE (mm): 556.8 -> 647.2 (BinsFormer depth) | Hurt |
| replace Depth Anything with fine-tuned BinsFormer for depth (ablation) | ViTPose + DETR · Rosetum3D 3D rose localization | MPJPE (mm): 656.9 -> 647.2 | Improved |
| compare detectors YOLOv5 vs DETR family | object detectors · Rosetum3D 2D detection | AP: 35.0 (YOLOv5) vs 33.0 (DETR) AP; AP50 67.1 vs 66.5 | Mixed |
Closest papers by meaning
- Reconstruction Using the Invisible: Intuition from NIR and Metadata for Enhanced 3D Gaussian Splatting (AAAI 2026 · similarity 0.84)
- Neural Hierarchical Decomposition for Single Image Plant Modeling (CVPR 2025 · similarity 0.83)
- Combining Deep Learning and Street View Imagery to Map Smallholder Crop Types (AAAI 2024 · similarity 0.82)
- AgMMU: A Comprehensive Agricultural Multimodal Understanding Benchmark (NeurIPS 2025 · similarity 0.81)
- Birth and Death of a Rose (CVPR 2025 · similarity 0.81)
- Can Large Multimodal Models Understand Agricultural Scenes? Benchmarking with AgroMind (NeurIPS 2025 · similarity 0.81)
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.
