Research map / Detection and segmentation / Open-set and few-shot detection
iLab-20M: A Large-Scale Controlled Object Dataset to Investigate Deep Learning
Check 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 iLab-20M rather than natural images for training | AlexNet fc7 SVM, boat vs. tank · Natural images | accuracy: 95 -> 78 | Hurt |
| Use iLab-20M rather than natural images for training | AlexNet fc7 SVM, boat vs. tank · iLab-20M | accuracy: 75 -> 97 | Helped |
| Fine-tune AlexNet on iLab-20M (ablation) | AlexNet fc7 SVM, boat vs. tank · Natural images | accuracy: 95 -> 93 | Hurt |
| Fine-tune AlexNet on iLab-20M (ablation) | AlexNet fc7 SVM, boat vs. tank · iLab-20M | accuracy: 97 -> 100 | Helped |
| Fine-tune AlexNet on iLab-20M (ablation) | AlexNet, seven-category classification · iLab-20M | accuracy: 92.5% -> 99.9% | Helped |
| Sample images randomly rather than systematically across one instance (ablation) | SVM category prediction with AlexNet fc7 · Four-category iLab-20M classification | accuracy | Helped |
| Sample images randomly rather than systematically across one instance (ablation) | SVM parameter prediction with AlexNet pool5 · Four-category iLab-20M classification | accuracy | Helped |
| Use pool5 rather than fc7 features (ablation) | AlexNet parameter prediction · iLab-20M | accuracy | Helped |
| Use fc7 rather than pool5 features (ablation) | AlexNet object-category prediction · iLab-20M | accuracy | Helped |
| Train on rotation labels before camera-view labels (ablation) | AlexNet pool5 parameter prediction · Camera-view prediction | accuracy: 77.05% (1.18) -> 89.20% (1.47) | Helped |
| Train on rotation labels before camera-view labels (ablation) | AlexNet pool5 parameter prediction · Rotation prediction | accuracy: 95.30% (1.00) -> 93.75%(1.66) | Hurt |
| Train on iLab-20M and test on natural scenes instead of matching the training and test domains | AlexNet fc7 SVM · Natural images | accuracy: 95 -> 78 | Hurt |
Closest papers by meaning
- Exploiting Invariance in Training Deep Neural Networks (AAAI 2022 · similarity 0.88)
- Progress and Limitations of Deep Networks to Recognize Objects in Unusual Poses (AAAI 2023 · similarity 0.88)
- Contemplating Real-World Object Classification (ICLR 2021 · similarity 0.87)
- Grounding inductive biases in natural images: invariance stems from variations in data (NeurIPS 2021 · similarity 0.87)
- ObjectNet: A large-scale bias-controlled dataset for pushing the limits of object recognition models (NeurIPS 2019 · similarity 0.87)
- Dataset Distillation for Pre-Trained Self-Supervised Vision Models (NeurIPS 2025 · similarity 0.86)
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.