Research map / Label-efficient and robust learning / Recognition and long-tail
MuCaLe-Net: Multi Categorical-Level Networks to Generate More Discriminating Features
Youssef Tamaazousti, Herve Le Borgne, Celine Hudelot
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 |
|---|---|---|---|
| combine features from subordinate- and basic-level networks (ablation) | MuCaLe-Net · VOC07 and NWO | mAP: 70.3 -> 72.5; 51.2 -> 54.1 | Improved |
| add a superordinate-label network to subordinate and basic-level networks (ablation) | MuCaLe-Net · VOC07 and NWO | mAP: 72.5 -> 73.0; 54.1 -> 54.9 | Improved |
| use only basic-level labels instead of subordinate labels (ablation) | MuCaLe-Net · VOC07 and NWO | mAP: 70.3 -> 70.0; 51.2 -> 51.0 | Hurt |
| use only superordinate labels instead of subordinate labels (ablation) | MuCaLe-Net · VOC07 and NWO | mAP: 70.3 -> 58.9; 51.2 -> 37.4 | Hurt |
| train an additional network from a different random initialization | standard CNN descriptor · VOC07 | mAP: 70.3 -> 70.5±0.3 | No clear effect |
| randomly group specific categories for generic labels | standard CNN descriptor · NWO | mAP: 51.2 -> 51.0±0.3 | Hurt |
| jointly train on specific and generic labels | standard CNN descriptor · VOC07 and NWO | mAP: 70.3 -> 71.5; 51.2 -> 53.4 | Improved |
| use independently trained categorical-level networks and concatenate their features | standard CNN descriptor · VOC07 and NWO | mAP: 70.3 -> 72.5; 51.2 -> 54.1 | Improved |
| use multi-categorical-level features | VGG-16 transfer representation · Caltech-101, Caltech-256, VOC 2007, VOC 2012 | accuracy or mAP: 88.8 -> 92.0; 78.0 -> 80.9; 86.1 -> 87.5; 84.5 -> 86.1 | Improved |
Closest papers by meaning
- Network of Experts for Large-Scale Image Categorization (ECCV 2016 · similarity 0.88)
- Learning from Small Sample Sets by Combining Unsupervised Meta-Training with CNNs (NeurIPS 2016 · similarity 0.86)
- Zero-Shot Recognition via Semantic Embeddings and Knowledge Graphs (CVPR 2018 · similarity 0.86)
- L-WISE: Boosting Human Visual Category Learning Through Model-Based Image Selection and Enhancement (ICLR 2025 · similarity 0.86)
- SPTNet: An Efficient Alternative Framework for Generalized Category Discovery with Spatial Prompt Tuning (ICLR 2024 · similarity 0.85)
- PICNN: A Pathway towards Interpretable Convolutional Neural Networks (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.
