-
Thanks for your code. But after I replaced fpn with your bifpn in libra rcnn, I got lower map. Did you get higher map in faster rcnn?
-
Hi,
when starting to train it produces the following error:
Error] Traceback (most recent call last):
File "train.py", line 226, in train
cls_loss, reg_loss = model(imgs, annot, obj_list=p…
-
on NVIDIA H800
I encountered a problem as shown in the title
here is my installing order:
0.cd SSIS
1.create a env (conda create -n ssis python=3.9) conda activate ssis
2.pip install onnx pro…
-
@zylo117 Thanks for your great work!
I'm now training EfficientDet-D0 on COCO2017 dataset with EfficientNet pretrained weights(https://github.com/lukemelas/EfficientNet-PyTorch). After around 100 epo…
-
**Short Description**
EfficientDet is a type of **object detection** model, which utilizes several optimization and backbone tweaks, such as the use of a [BiFPN](https://paperswithcode.com/method/b…
innat updated
2 months ago
-
EfficientDet: Scalable and Efficient Object Detection
* paper: https://arxiv.org/abs/1911.09070v1
> First, we propose a weighted bi-directional feature pyramid network (BiFPN), which allows ea…
-
I found it for a long time, but I didn't find out how to train. Is there no train code and I need to write it myself? Or I didn't find the train interface, If so, can you please give a train and valid…
-
![image](https://user-images.githubusercontent.com/88072901/157285133-3d1cb593-8176-4882-8c6c-c5f8e33ac855.png)
这个什么情况呢? 我就是把neck换成 bifpn 训练后 想导出训练的 模型 就提示这个错误
-
比如说我想把ppyolo中的yolofpn更换为BIFPN(其在modeling文件中已存在),应该修改哪些yml文件?
谢谢~
-
https://github.com/google/automl/blob/master/efficientdet/efficientdet_arch.py