Ma Shuhao

感觉这份代码对于实际工作几乎没什么意义

2

在eval结果中，要么只包含实例信息，要么只包含语义信息。那么实例信息中每个box属于什么语义并不清楚。而语义信息中的bounding box却是一个大的Box。也就是说每个实例的box没有与语义信息关联

I have read your papers(Understanding Convolution for semantic segmentation ),and run your net.Then I make an improvement.But When submit my results to pascal voc server,I always get an evaluation error....

ModuleNotFoundError: No module named 'crowdposetools'

4

ModuleNotFoundError: No module named 'crowdposetools'

在判断一个点是否落于三角行内，当出现如下情形： variable1 = np.array([[0, 20], [20, 20], [20, 20]]) variable2=np.array([x,x])时， isPointInTri(variable2， variable1) 这个函数的判断结果都为True, 也就是说，无论x为何值，结果都是True

I use the final.ckpt to inference the small dataset, and I set the argv --voxel_dim 208 208 48. But the result is a green map.Why? And I save the result...

`shape_loss = torch.sum(self.shape_prior(body_model_output.betas)) * self.shape_weight ** 2` So, what is the reason that the shape_loss. Anybody can explain it for me.

你好。我训练COCO数据集，发现先关键点loss居高不下

9

你好，我用您的代码训练COCOC数据集，发现'loss_classifier'会下降，但是 'loss_keypoint'基本居高不下

I think there is a trouble code

6

Please see the function that come from utils.py: def one_hot_it_v11_dice(label, label_info): semantic_map = [] void = np.zeros(label.shape[:2]) for index, info in enumerate(label_info): color = label_info[info][:3] class_11 = label_info[info][3] if class_11...

Is the receivers::Delaunay_grid function very very slow?

7

**NoiseModelling version** 4.0.2 **Describe the bug** when I use the receivers::Delaunay_grid, I found it is very very slow **To Reproduce** Steps to reproduce the behavior: 1. load BUILDINGS, ROADS and...

How to modify the shape value

1

### What's your question? I want to modify the shape value directly, by using sf.shape(i).points = [[xx,xx]]，but it is nothing useful