Research map / Detection and segmentation / CNN features
Deep Feature Factorization For Concept Discovery
Edo Collins, Radhakrishna Achanta, Sabine Susstrunk
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 |
|---|---|---|---|
| add CRF post-processing to DFF (ablation) | DFF · PASCAL VOC 2007 co-localization | CorLoc (%): 42.87 -> 43.51 | Improved |
| add CRF post-processing to DFF (ablation) | DFF (k=1) · iCoseg Elephants object co-segmentation | IoU (%): 65 -> 76 | Improved |
| add CRF post-processing to DFF (ablation) | DFF (k=1) · iCoseg Taj Mahal object co-segmentation | IoU (%): 41 -> 51 | Improved |
| add CRF post-processing to DFF (ablation) | DFF (k=1) · iCoseg Pyramids object co-segmentation | IoU (%): 57 -> 70 | Improved |
| add CRF post-processing to DFF (ablation) | DFF (k=1) · iCoseg Gymnastics1 object co-segmentation | IoU (%): 43 -> 52 | Improved |
| add CRF post-processing to DFF (ablation) | DFF (k=1) · iCoseg Statue of Liberty object co-segmentation | IoU (%): 49 -> 62 | Improved |
| add CRF post-processing to DFF (ablation) | DFF (k=3) · PASCAL-Parts cow head | avg IoU (%): 40.53 -> 45.20 | Improved |
| add CRF post-processing to DFF (ablation) | DFF (k=3) · PASCAL-Parts horse head | avg IoU (%): 28.85 -> 31.05 | Improved |
| add CRF post-processing to DFF (ablation) | DFF (k=3) · PASCAL-Parts cow neck+torso | avg IoU (%): 59.48 -> 58.87 | Hurt |
| add CRF post-processing to DFF (ablation) | DFF (k=3) · PASCAL-Parts horse neck+torso | avg IoU (%): 54.77 -> 53.18 | Hurt |
| use unsupervised DFF instead of best prior co-localization method (Le V) | Le (V) · PASCAL VOC 2007 co-localization | CorLoc (%): 41.97 -> 42.87 | Improved |
| use ResNet-101 instead of VGG family (ablation) | DFF · iCoseg | average IoU | Hurt |
Closest papers by meaning
- Concept Mask: Large-Scale Segmentation from Semantic Concepts (ECCV 2018 · similarity 0.86)
- Local Concept Embeddings for Analysis of Concept Distributions in Vision DNN Feature Spaces (IJCV 2025 · similarity 0.86)
- ICE: Intrinsic Concept Extraction from a Single Image via Diffusion Models (CVPR 2025 · similarity 0.84)
- Show and Tell: Visually Explainable Deep Neural Nets via Spatially-Aware Concept Bottleneck Models (CVPR 2025 · similarity 0.84)
- FACE: Faithful Automatic Concept Extraction (NeurIPS 2025 · similarity 0.83)
- Coarse-to-Fine Concept Bottleneck Models (NeurIPS 2024 · similarity 0.83)
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.
