Research map / Detection and segmentation / CNN features
MaskLab: Instance Segmentation by Refining Object Detection With Semantic and Direction Features
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 both semantic and direction features instead of semantic features alone (ablation) | MaskLab · COCO minival | mAP@0.75: 24.44% -> 29.72% | Helped |
| Add direction features to semantic features (ablation) | MaskLab · COCO minival | mAP@0.75: 24.44% -> 29.72% | Helped |
| Add 4-bin distance quantization to direction pooling (ablation) | MaskLab with semantic and direction features · COCO minival | mAP@0.75: 29.72% -> 30.57% | Helped |
| Use 8 directions rather than 6 (ablation) | MaskLab with 4 distance bins · COCO minival | mAP@0.75: 34.86% -> 34.82% | No clear effect |
| Add conv1 features to mask refinement (ablation) | MaskLab · COCO minival | mAP@0.75: 30.57% -> 32.92% | Helped |
| Add conv2 features alongside conv1 (ablation) | MaskLab mask refinement · COCO minival | mAP@0.75: 32.92% -> 33.89% | Helped |
| Add conv3 features to conv1 and conv2 (ablation) | MaskLab mask refinement · COCO minival | mAP@0.75: 33.89% -> 32.88% | Hurt |
| Initialize the box classifier with ImageNet pretrained weights (ablation) | MaskLab · COCO minival | mAP@0.75: 33.89% -> 34.82% | Helped |
| Pretrain the model on COCO semantic segmentation annotations (ablation) | MaskLab · COCO minival | mAP@0.75: 34.82% -> 35.91% | Helped |
| Use multi-grid atrous rates (ablation) | MaskLab · COCO minival | mAP@0.75: 35.91% -> 36.65% | Helped |
| Use random scale augmentation during training (ablation) | MaskLab · COCO minival | mAP@0.75: 38.61% -> 40.41% | Helped |
| Use output_stride=16 rather than output_stride=8 (ablation) | MaskLab · COCO minival | mAP@0.75: 40.41% -> 38.61% | Hurt |
Closest papers by meaning
- SIM: Semantic-Aware Instance Mask Generation for Box-Supervised Instance Segmentation (CVPR 2023 · similarity 0.90)
- Per-Pixel Classification is Not All You Need for Semantic Segmentation (NeurIPS 2021 · similarity 0.90)
- Mask DINO: Towards a Unified Transformer-Based Framework for Object Detection and Segmentation (CVPR 2023 · similarity 0.89)
- Boundary-Aware Instance Segmentation (CVPR 2017 · similarity 0.89)
- Fully Convolutional Instance-Aware Semantic Segmentation (CVPR 2017 · similarity 0.89)
- MaskDiff: Modeling Mask Distribution with Diffusion Probabilistic Model for Few-Shot Instance Segmentation (AAAI 2024 · similarity 0.87)
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.