Research map / Detection and segmentation / Tracking
Learning feed-forward one-shot learners
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 a single-stream learnet instead of a shared siamese network with weighted ℓ1 distance | Omniglot one-shot character recognition · Omniglot evaluation alphabets | error rate: 37.3% -> 28.6% | Helped |
| Use a weighted ℓ1 distance rather than a Euclidean distance in the single-stream learnet (ablation) | Learnet · Omniglot evaluation alphabets | error rate: 36.7 -> 28.6 | Helped |
| Use a dot product rather than weighted ℓ1 distance in the single-stream learnet (ablation) | Learnet · Omniglot evaluation alphabets | error rate: 28.6 -> 43.7 | Hurt |
| Factorize the convolutional layer (ablation) | Omniglot character recognition · Omniglot evaluation alphabets | error rate | No clear effect |
| Use distinct parameters for the two streams instead of shared parameters (ablation) | Siamese network with weighted ℓ1 distance · Omniglot evaluation alphabets | error rate: 41.8 -> 34.6 | Helped |
| Replace static filters with learnet-predicted filters | Siamese tracker (φ=B) · VOT 2015 | accuracy: 0.465 -> 0.500 | Helped |
| Factorize the filters without using a learnet (ablation) | Siamese tracker (φ=B) · VOT 2015 | accuracy: 0.465 -> 0.444 | Hurt |
| Use distinct parameters for the two streams instead of shared parameters (ablation) | Siamese tracker (φ=B) · VOT 2015 | accuracy: 0.465 -> 0.447 | Hurt |
| Use a slimmer prediction network B rather than A (ablation) | Siamese learnet tracker (φ=B) · VOT 2015 | accuracy: 0.500 -> 0.497 | No clear effect |
| Use factorized convolutions without a learnet (ablation) | Siamese tracker (φ=C) · VOT 2015 | accuracy: 0.466 -> 0.435 | Hurt |
| Use a slimmer prediction network C rather than A (ablation) | Siamese learnet tracker (φ=C) · VOT 2015 | accuracy: 0.483 -> 0.491 | Helped |
| Use the learnet tracker rather than Siamese (φ=B) | Visual object tracking · VOT 2015 | failures: 105 -> 87 | Helped |
Closest papers by meaning
- Matching Networks for One Shot Learning (NeurIPS 2016 · similarity 0.90)
- ‘Less Than One’-Shot Learning: Learning N Classes From M < N Samples (AAAI 2021 · similarity 0.88)
- Multi-Attention Network for One Shot Learning (CVPR 2017 · similarity 0.87)
- CLEAR: Cumulative LEARning for One-Shot One-Class Image Recognition (CVPR 2018 · similarity 0.86)
- One-Shot Image Classification by Learning to Restore Prototypes (AAAI 2020 · similarity 0.86)
- Few-shot Learner Parameterization by Diffusion Time-steps (CVPR 2024 · 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.