Research map / Continual learning / Class-incremental
Data-Free Class-Incremental Hand Gesture Recognition
Shubhra Aich, Jesus Ruiz-Santaquiteria, Zhenyu Lu, Prachi Garg, K J Joseph, Alvaro Fernandez Garcia, Vineeth N Balasubramanian, Kenrick Kin, Chengde Wan, Necati Cihan Camgoz, Shugao Ma, Fernando De la Torre
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 |
|---|---|---|---|
| invert boundary-aware prototypical samples plus support vectors instead of using existing DFCIL methods | DG-STA gesture recognizer · SHREC-2017 mean Task 1-6 | global accuracy G (%): 56.2 -> 70.5 vs ABD | Improved |
| apply BOAT-MI on EgoGesture3D | DG-STA gesture recognizer · EgoGesture3D mean Task 1-6 | global accuracy G (%): 52.6 -> 61.2 vs ABD | Improved |
| invert boundary samples instead of random samples (ablation) | synthetic MLP · Synthetic Gaussian blobs | total accuracy (%): 58.6 -> 90.8 | Improved |
| sample at a small margin inside boundary instead of exactly on it (ablation) | prototypical MI · SHREC-2017 | IFM (%): 15.1 -> 10.5 | Improved |
| sample deep inside class (delta 0.8) rather than near boundary (ablation) | prototypical MI · SHREC-2017 Task 6 | global accuracy G (%): 53.4 -> 50.1 vs near-boundary | Hurt |
| use features at random distances instead of margin-based sampling (ablation) | prototypical MI · SHREC-2017 | global accuracy G (%): 69.5 -> 66.0 | Hurt |
| add prototypical boundary samples to standard SVs (ablation) | BOAT-MI · SHREC-2017 | global accuracy G (%): 61.4 -> 70.5 | Improved |
| add SVs to prototypical samples (ablation) | BOAT-MI · SHREC-2017 | IFM (%): 10.8 -> 8.8 | Improved |
| use data-free inversion methods versus simple feature extraction baseline | DFCIL methods ABD, R-DFCIL, DeepInversion · SHREC-2017 mean Task 1-6 | global accuracy G (%): feature extraction 60.6 vs ABD 56.2 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- R-DFCIL — R-DFCIL: Relation-Guided Representation Learning for Data-Free Class Incremental Learning (ECCV 2022) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Decoupled Representation Learning for Skeleton-Based Gesture Recognition (CVPR 2020 · similarity 0.87)
- SeqHAND: RGB-Sequence-Based 3D Hand Pose and Shape Estimation (ECCV 2020 · similarity 0.85)
- Online Detection and Classification of Dynamic Hand Gestures With Recurrent 3D Convolutional Neural Network (CVPR 2016 · similarity 0.85)
- FreiHAND: A Dataset for Markerless Capture of Hand Pose and Shape From Single RGB Images (ICCV 2019 · similarity 0.85)
- Collaborative Learning of Gesture Recognition and 3D Hand Pose Estimation with Multi-Order Feature Analysis (ECCV 2020 · similarity 0.85)
- Hand-Transformer: Non-Autoregressive Structured Modeling for 3D Hand Pose Estimation (ECCV 2020 · similarity 0.85)
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.
