Research map / Video / Video segmentation and tracking
Privacy-Preserving Human Activity Recognition from Extreme Low Resolution
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use ISR-generated training samples with learned transformations | PoT, HOG + HOF + CNN · DogCentric 16x12 | accuracy: 64.6 % -> 67.4 % | Helped |
| Use ISR-generated training samples | PoT · JPL-Interaction 16x12 | accuracy: 92.9 % -> 96.4 % | Helped |
| Learn transformations by decision-boundary matching rather than using the resized baseline | PoT, DogCentric 16x12, linear kernel · DogCentric 16x12 | accuracy: 58.47 -> 61.73 | Helped |
| Learn transformations by maximum entropy rather than using the resized baseline | PoT, DogCentric 16x12, linear kernel · DogCentric 16x12 | accuracy: 58.47 -> 61.96 | Helped |
| Use maximum-entropy transform learning instead of random-transform data augmentation | PoT, DogCentric 16x12, linear kernel · DogCentric 16x12 | accuracy: 60.86 -> 61.96 | Helped |
| Add random rotations to data augmentation | PoT, DogCentric 16x12 · DogCentric 16x12 | accuracy: 60.86 -> 60.94; 63.36 -> 64.17; 62.08 -> 62.85 | Mixed |
| Choose transformations uniformly instead of learning them with ISR | PoT, DogCentric 16x12 · DogCentric 16x12 | accuracy: 61.73 -> 60.56; 64.85 -> 63.95; 63.35 -> 62.29 | Hurt |
| Use maximum-entropy transform learning instead of decision-boundary matching | PoT, DogCentric 16x12 · DogCentric 16x12 | accuracy: 61.73 -> 61.96; 64.85 -> 64.91; 63.35 -> 63.61 | Mixed |
| Generate additional ISR samples | ISR-method2 · HMDB 16x12 and 32x24 | accuracy | Helped |
| Use 32x24 rather than 16x12 videos | ISR-method2 · HMDB | accuracy | Helped |
Closest papers by meaning
- Extreme Low Resolution Activity Recognition With Multi-Siamese Embedding Learning (AAAI 2018 · similarity 0.89)
- Privacy Beyond Pixels: Latent Anonymization for Privacy-Preserving Video Understanding (ICLR 2026 · similarity 0.85)
- SPAct: Self-Supervised Privacy Preservation for Action Recognition (CVPR 2022 · similarity 0.84)
- ReGenHuman: Re-Generating Human Appearances for Realistic Full-Body Video Anonymization (NeurIPS 2026 · similarity 0.83)
- HARDVS: Revisiting Human Activity Recognition with Dynamic Vision Sensors (AAAI 2024 · similarity 0.83)
- Selective Interpretable and Motion Consistent Privacy Attribute Obfuscation for Action Recognition (CVPR 2024 · 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.