1. We find that the performance is unstable with 1x setting and may fluctuate by about 0.3 mAP. mAP 40.3 ~ 40.6 is acceptable. Such fluctuation can also be found in the original implementation.
2. You can get a more stable results ~ mAP 40.6 with a schedule total 13 epoch, and learning rate is divided by 10 at 10th and 13th epoch.
We provide config files to reproduce the object detection & instance segmentation results in the ICCV 2019 Oral paper for [CARAFE: Content-Aware ReAssembly of FEatures](https://arxiv.org/abs/1905.02188).
```
@inproceedings{Wang_2019_ICCV,
title = {CARAFE: Content-Aware ReAssembly of FEatures},
author = {Wang, Jiaqi and Chen, Kai and Xu, Rui and Liu, Ziwei and Loy, Chen Change and Lin, Dahua},
booktitle = {The IEEE International Conference on Computer Vision (ICCV)},
month = {October},
year = {2019}
}
```
## Results and Models
The results on COCO 2017 val is shown in the below table.
| Method | Backbone | Style | Lr schd | Test Proposal Num | Inf time (fps) | Box AP | Mask AP | Config | Download |
