Research map / Detection and segmentation / CNN features
YOLACT: Real-Time Instance Segmentation
Daniel Bolya, Chong Zhou, Fanyi Xiao, Yong Jae Lee
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 Fast NMS in place of standard NMS (ablation) | YOLACT-550 · COCO val2017 | mask AP: 30.0 -> 29.9 | No clear effect |
| add Fast NMS in place of standard NMS (ablation) | YOLACT-550 · COCO val2017 | FPS: 24.0 -> 33.5 | Improved |
| add Fast NMS in place of standard NMS (ablation) | Mask R-CNN · COCO val2017 | mask AP: 36.1 -> 35.8 | Hurt |
| add semantic segmentation loss during training (ablation) | YOLACT-550 · COCO | mask AP: +0.4 mAP | Improved |
| increase number of prototypes from 8 to 32 (ablation) | YOLACT-550 · COCO val2017 | mask AP: 26.8 -> 27.7 | Improved |
| increase prototypes from 32 to 64 (ablation) | YOLACT-550 · COCO val2017 | mask AP: 27.7 -> 27.8 | No clear effect |
| increase prototypes from 64 to 256 (ablation) | YOLACT-550 · COCO val2017 | mask AP: 27.8 -> 27.7 | No clear effect |
| add prototype plus coefficient mask assembly instead of fc mask output (ablation) | one-stage detector · COCO val2017 | mask AP: 20.7 -> 29.9 | Improved |
| increase image size from 400 to 550 (ablation) | YOLACT · COCO test-dev | mask AP: 24.9 -> 29.8 | Improved |
| increase image size from 550 to 700 (ablation) | YOLACT · COCO test-dev | mask AP: 29.8 -> 31.2 | Improved |
| use ResNet-101 instead of ResNet-50 backbone (ablation) | YOLACT-550 · COCO test-dev | mask AP: 28.2 -> 29.8 | Improved |
| use DarkNet-53 instead of ResNet-50 backbone (ablation) | YOLACT-550 · COCO test-dev | mask AP: 28.2 -> 28.7 | Improved |
Later papers that compared with this method
5 later accepted papers report YOLACT-700 as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- Instance and Panoptic Segmentation Using Conditional Convolutions (TPAMI 2022 · better on 2 of 2 shared results)
- SOTR: Segmenting Objects With Transformers (ICCV 2021 · better on 1 of 2 shared results)
- BoxInst: High-Performance Instance Segmentation With Box Annotations (CVPR 2021 · better on 1 of 1 shared result)
- CenterMask: Real-Time Anchor-Free Instance Segmentation (CVPR 2020 · better on 1 of 1 shared result)
- Deep Variational Instance Segmentation (NeurIPS 2020 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- PA-Net — PANet: Few-Shot Image Semantic Segmentation With Prototype Alignment (ICCV 2019) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- YOLACT++ Better Real-Time Instance Segmentation (TPAMI 2020 · similarity 0.97)
- FastInst: A Simple Query-Based Model for Real-Time Instance Segmentation (CVPR 2023 · similarity 0.91)
- SOLO: A Simple Framework for Instance Segmentation (TPAMI 2021 · similarity 0.90)
- SOLO: Segmenting Objects by Locations (ECCV 2020 · similarity 0.90)
- BlendMask: Top-Down Meets Bottom-Up for Instance Segmentation (CVPR 2020 · similarity 0.90)
- Real-Time Instance Segmentation With Discriminative Orientation Maps (ICCV 2021 · similarity 0.89)
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.
