-
### Issue Type
Bug
### OS
Ubuntu
### OS architecture
x86_64
### Programming Language
Other
### Framework
Other
### Model name and Weights/Checkpoints URL
228_Fast-SCNN
https://github.com/P…
-
### Detailed Description of the Problem
OpenShift 4 is currently using haproxy 2.2. We really, really want to
move on to the 2.6 series however we have identified what appears to
be either differ…
-
I'm trying for a couple of days now to make use of this repo, but I can't make it work...
[Here](https://colab.research.google.com/drive/1L0Gq4qhPoGMVmJklX82vIu-mMBIvphDz?usp=sharing) is what I've tr…
-
The plantform is win10, I install mmcv-full 1.1.5 for mmseg. It works well when I run main.py in pycharm.
However, after using pyinstaller package it a exe. It wrongs! ModuleNotFoundError: No module …
-
[SKIP] Shape of pretrained weight ./pretrained_model/fast_scnn_ci…
-
采用`Fast-scnn`进行训练。训练完后转化成静态图效果骤降。
如下图所示，最左侧为原图，中间为动态图推理，右侧为静态图推理。
如图，我用其他模型如faster-scnn和ocrnet去训练都没有这个错误。
唯独u2net不行，conf…
-
We would like to add main figure and abstract of each model in our repo. This issue is used for uploading figures for hyperlink in README.md
-
使用paddleSeg里的cityscape_fast_scnn.yaml，GPU利用率很低，batch_size是32，显存利用率50%多，但是GPU利用率只有不到5%.尝试过在yaml文件里更改DATALOADER: NUM_WORKERS: 320 BUF_SIZE: 2560，但是没有一点作用
-
我按照教程训练制作自己的数据集，使用Fast scnn 进行多分类语义分割训练，配置文件也是按照教程来的，但是训练结果只是二分类，请问是什么原因。
train_dataset:
type: Dataset
dataset_root: data
train_path: data/train.txt
num_classes: 25
transforms:
- …