-
Thanks for sharing the code. I am wondering why not using Softmax+cross_entropy?
Did you see any differences in performance?
-
I am running the repo on RTX2060 with pytorch1.10 + cuda 10.2 environment and got this error
File "C:\Users\Ali\anaconda3\envs\detr\lib\site-packages\portalocker-2.4.0-py3.10.egg\portalocker\portaloc…
-
你好，paddle版本有没有可视化特征图代码呢？
网上一堆pytorch的，我用的paddle框架，啥可视化都画不出呜呜，现在来不及换pytorch版本
-
论文中decoder的消融实验介绍模型可以无需重新训练而使用不同的解码器层，请问一下我需要调整哪里的参数来更换decoder层数？
我在 num_decoder_layers直接更换decoder层数时验证模型会报错。
的模型的backbone是用的eva02的det任务给出的eva02_L_pt_m38m_p14to16 | 304M | Merged-38M | 56这个预训练模型嘛，如果不是是否可以说明一下具体使用的模型。论文中提到的`Co-DINO-Deformable-DETR`各部分是如何组成的呢，感谢！
-
## 🚀 Feature
Adding new models to the models section.
## Motivation
Many new models have been proposed in the recent years and do not exist in the models module.
For example, the EfficientNets…
-
https://github.com/fundamentalvision/Deformable-DETR/blob/11169a60c33333af00a4849f1808023eba96a931/models/segmentation.py#L221
The focal loss is computed on all object queries (i.e. suppose `num_qu…
-
-
## 🚀 Feature
Frequently re-used losses that can be added.
### Motivation
Writing Losses is quite repetitive. PyTorch supports losses which are written with deep interoperability with C++ API. …