Research map / Video / Video segmentation and tracking
MotionDeltaCNN: Sparse CNN Inference of Frame Differences in Moving Camera Videos with Spherical Buffers and Padded Convolutions
Mathias Parger, Chengcheng Tang, Thomas Neff, Christopher D. Twigg, Cem Keskin, Robert Wang, Markus Steinberger
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 dynamic initialization of newly unveiled tiles (ablation) | MotionDeltaCNN · DAVIS 2017 moving-camera, 1080 Ti b=5 | FPS: 56.4 -> 75.1 FPS | Improved |
| add dynamic initialization of newly unveiled tiles (ablation) | MotionDeltaCNN · DAVIS 2017 moving-camera | IoU: 79.63 -> 79.43 | Mixed |
| add padded convolutions (ablation) | MotionDeltaCNN · DAVIS 2017 moving-camera | IoU: 77.89 -> 79.43 | Improved |
| add ROI-focused sensing (ablation) | MotionDeltaCNN · DAVIS 2017 moving-camera | FPS: 63.6 -> 75.1 FPS | Improved |
| add noise suppression and ROI scaling (ablation) | MotionDeltaCNN · DAVIS 2017 moving-camera | FPS: 50.9 -> 75.1 FPS | Improved |
| add moving-camera support (spherical buffers, padded convs) to DeltaCNN | DeltaCNN sparse · DAVIS 2017 moving-camera videos, 1080 Ti b=5 | FPS: 44.1 -> 75.1 FPS | Improved |
| use MotionDeltaCNN instead of DeltaCNN on static camera Human3.6M | HRNet pose estimation · Human3.6M, 1080 Ti b=16 | FPS: 202 -> 228 FPS | Improved |
| run MotionDeltaCNN on RTX 3090 with small batch size | MotionDeltaCNN · DAVIS 2017, RTX 3090 b=1 | speedup over cuDNN: 1.44 (MotionDeltaCNN) vs 1.54 (DeltaCNN) | Mixed |
| enable moving-camera support in dense mode | DeltaCNN (dense) · DAVIS 2017 all videos, 1080 Ti b=5 | speedup over cuDNN: 1.01 -> 0.99 | No clear effect |
Closest papers by meaning
- DeltaCNN: End-to-End CNN Inference of Sparse Frame Differences in Videos (CVPR 2022 · similarity 0.91)
- SDC-Net: Video prediction using spatially-displaced convolution (ECCV 2018 · similarity 0.87)
- Click To Move: Controlling Video Generation With Sparse Motion (ICCV 2021 · similarity 0.86)
- Learning to Segment Moving Objects (IJCV 2018 · similarity 0.85)
- Learning Motion Patterns in Videos (CVPR 2017 · similarity 0.85)
- TTVidT: Decoupling the Temporal Axis for Efficient Motion-Centric Video Pretraining (NeurIPS 2026 · similarity 0.84)
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.
