boxlevelset issues

tree_filter

Thank you very much for your work, but tree_filter is very hard to install, do you have another implementation or another option available?

Wenlve-Zhou

您好，我最近想用这个方法做一个输电线路数据集的弱监督分割，分割目标是输电线，只有一类，并且想要的到的评价指标为语义分割常用的mIOU和像素级的Precision，但boxlevelset中的评价指标为mAP值，我一步步查看了cocoapi的源码，发现cocoapi中的computeIou计算得到的是一个矩阵，我有尝试过将网络预测结果保存为json文件，在自己写一个代码计算像素级的mIOU和Precision，但保存的json文件为rle格式的，并且只能是这个格式。请问我怎么才能计算预测结果和标签的mIOU和Precision值？谢谢。

TaoshuaiZ

IndexError: list index out of range

5

I ran into this problem when training with a custom data set.How to solve it? Traceback (most recent call last): File "tools/train.py", line 191, in main() File "tools/train.py", line 187,...

ccdongxu

关于两个遥感数据集表现差异非常大的问题，loss_levelset在表现差的数据集上无法下降

3

![025735bb69a21084deba9f0d19e7500](https://github.com/LiWentomng/boxlevelset/assets/47242198/3dbfa362-82c2-4b46-8aa8-1a7ea651a661) 作者您好，非常感谢您提供的代码，我在自己的三个建筑物数据集上进行实验时，发现levelset的表现差异很大，loss_levelset这项损失在来同一个城市的两个数据集上无法下降，即使后来将batchsize设置为4，初始学习率设置为0.00125时都无法下降，因此在这两个数据集上的map相比全监督的方法低30％，即使是边界框的map也比全监督实例分割方法的边界框map低30％。而在来自另一个城市的数据集上精度可以和全监督的方法媲美。请问这可能是什么原因造成的呢？

lin1061991611

ValueError: need at least one array to concatenate while training on custom data

1

Hi, thanks for sharing this work. I would like to train the model on custom data and I have a data with coco annotation format and I have this error...

assia855

question

1

请问boxlevelset在mmdet框架中测试，如何只输出掩膜图？就是不需要有原图的叠加

Huster-Hq

Instance Segmentation on the Whole Image

1

@LiWentomng In your paper, you mentioned restricting the level set to gt bounding box for segmentation and also mentioned that the training result is poor on the whole image. How...

zhaoyangwei123

Instance and Panoptic segmentation

1

@LiWentomng thanks for thsi toolbox it shall be very helpfull to the community, just had few queries can we use this to train instance segmentation using BDD100k dataset for all...

abhigoku10