Research map / Video / Action recognition
Reconstruction-Free Action Inference from Compressive Imagers
Kuldeep Kulkarni, Pavan K. Turaga
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 affine-compensated filter matched to the test viewing angle (15 deg) instead of canonical filter (ablation) | STSF MACH filters · Weizmann simulated 15 deg view | recognition rate (%): 65.56 -> 72.22 | Improved |
| compensate filter for 5 deg view when test view is 15 deg (ablation) | STSF · Weizmann simulated 15 deg view | recognition rate (%): 65.56 -> 68.88 | Improved |
| compensate filter for 20 deg view when test view is 15 deg (mismatched) (ablation) | STSF · Weizmann simulated 15 deg view | recognition rate (%): 65.56 -> 66.67 | Mixed |
| operate on compressed measurements with STSF instead of reconstruct with CoSaMP then IDT, CR 100 | STSF vs Recon+IDT · Weizmann | recognition rate (%): 5.56 -> 81.11 | Improved |
| STSF instead of Recon+IDT at CR 500 | STSF vs Recon+IDT · Weizmann | recognition rate (%): 7.77 -> 78.89 | Improved |
| increase compression from 100 to 300 in STSF | STSF · Weizmann | recognition rate (%): 81.11 -> 76.66 | Hurt |
| STSF instead of Recon+IDT, CR 100 | STSF vs Recon+IDT · UCF50 | recognition rate (%): 21.72 -> 54.55 | Improved |
| STSF instead of Recon+IDT, CR 400 | STSF vs Recon+IDT · UCF50 | recognition rate (%): 12.52 -> 46.48 | Improved |
| STSF instead of Recon+IDT, CR 100 | STSF vs Recon+IDT · HMDB51 | recognition rate (%): 6.23 -> 21.125 | Improved |
| STSF instead of Recon+IDT, CR 400 | STSF vs Recon+IDT · HMDB51 | recognition rate (%): 2.33 -> 17.02 | Improved |
| compress at CR 100 versus uncompressed Oracle MACH | STSF · UCF Sports | recognition rate (%): 69.2 -> 70.67 | No clear effect |
| use full-blown images with Recon+IDT instead of compressed STSF (CR 1) | Recon+IDT · UCF50 | recognition rate (%): 60.86 -> 91.2 | Improved |
Closest papers by meaning
- Selective Feature Compression for Efficient Activity Recognition Inference (ICCV 2021 · similarity 0.87)
- Compressed Video Action Recognition (CVPR 2018 · similarity 0.85)
- Action Recognition in Video Using Sparse Coding and Relative Features (CVPR 2016 · similarity 0.85)
- Motion-Aware Dynamic Graph Neural Network for Video Compressive Sensing (TPAMI 2024 · similarity 0.85)
- DetectSCI: Toward Object-Guided ROI Reconstruction for High-Resolution Video Snapshot Compressive Imaging (CVPR 2026 · similarity 0.85)
- TrAction: Action Recognition with Sparse Trajectories (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.
