-
Thank you for your amazing job!
I read your paper and found that you mentioned Fully-connected FPN(**Path-aggregation blocks**) in the paper, but did not write the citation source. Could you give me…
-
Can not find the FPN part, could you please kindly point it out?
-
### Prerequisite
- [X] I have searched [Issues](https://github.com/open-mmlab/mmdetection3d/issues) and [Discussions](https://github.com/open-mmlab/mmdetection3d/discussions) but cannot get the expec…
-
## 🚀 Feature
Panoptic-FPN on Cityscapes
## Motivation
We want to use Panoptic-FPN as our baseline on Cityscapes benchmark.
## Pitch
We hope to access the whole training configs/scr…
-
您好,请问pytorch版本的graphFPN发布在什么地址了呢?
-
First of all thank you for releasing the code and ckpts.
I am not able to reproduce the results using config file "sparse_rcnn_focalnet_tiny_fpn_300_proposals_crop_mstrain_480-800_3x_coco_lrf.py" an…
-
For example, I have a faster-rcnn model pre-trained on coco dataset and I am fine-tuning on downstream dataset. To analyse the features of the detector, I want to extract the embeddings from the neck…
-
When I replace the FPN with ASFF in Retinaface, the model size is double, but the result is inferior to FPN.
-
- 负责人:
zhaoyuchen
- 测试环境
- 显卡驱动:418.39
- CUDA 9.0 CUDNN7
- 当前性能对比
CUDA90
backbone:resnext
场景 | Paddle | pytorch | 对比
---|---|---|---
单GPU | 2.601 | 3.035 | 差于 14%
backbon…
-
I want to use FPN for the VGG model since I only have 2 gtx 980 with 8 G memory, do you have plan to share a VGG based model ?