| Training Techniques | InstaBoost |
|---|---|
| Architecture | RPN, FPN, Cascade R-CNN, ResNet, RoIAlign |
| lr sched | 4x |
| SHOW MORE |
| Training Techniques | InstaBoost |
|---|---|
| Architecture | Softmax, RPN, Convolution, Dense Connections, FPN, ResNet, RoIAlign |
| lr sched | 4x |
| SHOW MORE |
| Training Techniques | InstaBoost |
|---|---|
| Architecture | Softmax, RPN, Convolution, Dense Connections, FPN, ResNet, RoIAlign |
| lr sched | 4x |
| SHOW MORE |
| Training Techniques | InstaBoost |
|---|---|
| Architecture | Softmax, RPN, ResNeXt, Convolution, Dense Connections, FPN, RoIAlign |
| lr sched | 4x |
| SHOW MORE |
[ALGORITHM]
Configs in this directory is the implementation for ICCV2019 paper "InstaBoost: Boosting Instance Segmentation Via Probability Map Guided Copy-Pasting" and provided by the authors of the paper. InstaBoost is a data augmentation method for object detection and instance segmentation. The paper has been released on arXiv.
@inproceedings{fang2019instaboost,
title={Instaboost: Boosting instance segmentation via probability map guided copy-pasting},
author={Fang, Hao-Shu and Sun, Jianhua and Wang, Runzhong and Gou, Minghao and Li, Yong-Lu and Lu, Cewu},
booktitle={Proceedings of the IEEE International Conference on Computer Vision},
pages={682--691},
year={2019}
}
You need to install instaboostfast before using it.
pip install instaboostfast
The code and more details can be found here.
InstaBoost have been already integrated in the data pipeline, thus all you need is to add or change InstaBoost configurations after LoadImageFromFile. We have provided examples like this. You can refer to InstaBoostConfig for more details.
coco_2017_train and tested on coco_2017_val for conveinience of evaluation and comparison. In the paper, the results are obtained from test-dev.| Network | Backbone | Lr schd | Mem (GB) | Inf time (fps) | box AP | mask AP | Config | Download |
|---|---|---|---|---|---|---|---|---|
| Mask R-CNN | R-50-FPN | 4x | 4.4 | 17.5 | 40.6 | 36.6 | config | model | log |
| Mask R-CNN | R-101-FPN | 4x | 6.4 | 42.5 | 38.0 | config | model | log | |
| Mask R-CNN | X-101-64x4d-FPN | 4x | 10.7 | 44.7 | 39.7 | config | model | log | |
| Cascade R-CNN | R-101-FPN | 4x | 6.0 | 12.0 | 43.7 | 38.0 | config | model | log |
Object Detection on COCO minival
| MODEL | BOX AP |
|---|---|
| Mask R-CNN InstaBoost (X-101-64x4d-FPN, 4x) | 44.7 |
| Cascade R-CNN InstaBoost (R-101-FPN, 4x) | 43.7 |
| Mask R-CNN InstaBoost (R-101-FPN, 4x) | 42.5 |
| Mask R-CNN InstaBoost (R-50-FPN, 4x) | 40.6 |
Instance Segmentation on COCO minival
| MODEL | MASK AP |
|---|---|
| Mask R-CNN InstaBoost (X-101-64x4d-FPN, 4x) | 39.7 |
| Cascade R-CNN InstaBoost (R-101-FPN, 4x) | 38.0 |
| Mask R-CNN InstaBoost (R-101-FPN, 4x) | 38.0 |
| Mask R-CNN InstaBoost (R-50-FPN, 4x) | 36.6 |
