Research map / Video / Video segmentation and tracking
Deep Hand: How to Train a CNN on 1 Million Hand Images When Your Data Is Continuous and Weakly Labelled
Oscar Koller, Hermann Ney, Richard Bowden
Check 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 1-Million-Hands features to the [18] multimodal system | [18] sign language recognition system · PHOENIX 2014 Dev | WER: 57.3 -> 47.1 | Improved |
| Add 1-Million-Hands features to the [18] multimodal system | [18] sign language recognition system · PHOENIX 2014 Test | WER: 55.6 -> 45.1 | Improved |
| Add 1-Million-Hands features to the [18] multimodal system | [18] sign language recognition system · SIGNUM Test | WER: 10.0 -> 7.6 | Improved |
| Replace HoG-3D features with 1-Million-Hands CNN features | Hand-only continuous sign language recognition · PHOENIX 2014 Dev | WER: 60.9 -> 51.6 | Improved |
| Replace HoG-3D features with 1-Million-Hands CNN features | Hand-only continuous sign language recognition · PHOENIX 2014 Test | WER: 58.1 -> 50.2 | Improved |
| Replace HoG-3D features with 1-Million-Hands CNN features | Hand-only continuous sign language recognition · SIGNUM Test | WER: 12.5 -> 12.0 | Improved |
| Add PHOENIX training data to the Danish and New Zealand data (ablation) | CNN hand-shape classifier · PHOENIX 2014 manually annotated hand shapes | top-1 accuracy: 59.6 -> 62.8; 59.5 -> 57.7 | Mixed |
| Train with Danish data only rather than adding New Zealand and PHOENIX data (ablation) | CNN hand-shape classifier · PHOENIX 2014 manually annotated hand shapes | top-1 accuracy: 50.6 -> 62.8 | Hurt |
| Train with Danish and New Zealand data rather than also adding PHOENIX data (ablation) | CNN hand-shape classifier · PHOENIX 2014 manually annotated hand shapes | top-1 accuracy: 59.6 -> 62.8; 59.5 -> 57.7 | Mixed |
Closest papers by meaning
- Weakly Supervised Learning with Multi-Stream CNN-LSTM-HMMs to Discover Sequential Parallelism in Sign Language Videos (TPAMI 2019 · similarity 0.91)
- Hand-Model-Aware Sign Language Recognition (AAAI 2021 · similarity 0.88)
- Deep Sign: Enabling Robust Statistical Continuous Sign Language Recognition via Hybrid CNN-HMMs (IJCV 2018 · similarity 0.88)
- Recurrent Convolutional Neural Networks for Continuous Sign Language Recognition by Staged Optimization (CVPR 2017 · similarity 0.88)
- Re-Sign: Re-Aligned End-To-End Sequence Modelling With Deep Recurrent CNN-HMMs (CVPR 2017 · similarity 0.88)
- A Wireframe-Based Approach for Classifying and Acquiring Proficiency in the American Sign Language (Student Abstract) (AAAI 2024 · 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.
