black region in generated labels

[gongshichina]

Thanks for your wonderful work and codes!

I used your scripts to generate the nuscens BEV labels, but it has some black regions in the drivable(green circle above), which have much bad effect in our performance evalutaion. I also have tried map-v1.0,-v1.2,-v1.3, and they all have the same problem. Could you please give me any advice? Many thanks again!

[songqi-github]

Hi, do you got any problems in visualizing the masked labels?

my visualized labels:

visualized labels in this paper:

It seems that my visualized results are over-masked compared with the picture in this paper.

[songqi-github]

The mask I used is generated by the following code in this repo:

    # Load label image as a torch tensor
    label_path = os.path.join('/data/full-v1.0/nuscenes/map-labels-v1.2', token + '.png')
    encoded_labels = to_tensor(Image.open(label_path)).long()
    # Decode to binary labels
    labels = decode_binary_labels(encoded_labels, NUM_CLASS + 1)
    labels, mask = labels[:-1], ~labels[-1]

[vonshun]

I have the same problems like the two above issues.T^T

[liqiokkk]

Hi, do you got any problems in visualizing the masked labels?

my visualized labels:

visualized labels in this paper:

It seems that my visualized results are over-masked compared with the picture in this paper.

Hi, Could you use it normally? I had the same problem as you in https://github.com/tom-roddick/mono-semantic-maps/issues/10.

[wang-TJ-20]

@songqi-github hello, I used scripts to generate the nuscens BEV labels, What I got is different from you. It's a grayscale image without anything.

[Estrellama]

I also have tried map-v1.0,-v1.2,-v1.3,

Hi，I used scripts to generate the nuscens BEV labels in map-v1.0 follow the repo, but error occurs:

Exception: Error: You are using an outdated map version (%s)! Please go to https://www.nuscenes.org/download to download the latest map!

How did you overcome the problem of the dataset version and use map-v1.0 to generate labels? Any advice？Thanks again!

[Drchz155]

感谢您的出色工作和代码！

我使用您的脚本生成了 nuscens BEV 标签，但它在可驾驶（上面的绿色圆圈）中有一些黑色区域，这对我们的性能评估有很多不良影响。我也尝试过map-v1.0，-v1.2，-v1.3，它们都有同样的问题。你能给我任何建议吗？再次感谢！

Thank you for your comment. Has your issue been resolved? I want to generate labels through the code in the project, but only get some grayscale images. Could you provide your related code, or compress and upload your labels? Thanks again.

[Drchz155]

感谢您的出色工作和代码！

我使用您的脚本生成了 nuscens BEV 标签，但它在可驾驶（上面的绿色圆圈）中有一些黑色区域，这对我们的性能评估有很多不良影响。我也尝试过map-v1.0，-v1.2，-v1.3，它们都有同样的问题。你能给我任何建议吗？再次感谢！

Thank you for your comment. Has your issue been resolved? I want to generate labels through the code in the project, but only get some grayscale images. Could you provide your related code, or compress and upload your labels? Thanks again.

[Drchz155]

@songqi-github hello, I used scripts to generate the nuscens BEV labels, What I got is different from you. It's a grayscale image without anything.

请问你的问题解决了么，我和你遇到了一样的问题

[Drchz155]

感谢您的出色工作和代码！

我使用您的脚本生成了 nuscens BEV 标签，但它在可驾驶（上面的绿色圆圈）中有一些黑色区域，这对我们的性能评估有很多不良影响。我也尝试过map-v1.0，-v1.2，-v1.3，它们都有同样的问题。你能给我任何建议吗？再次感谢！

您好，可以分享一下您的标签生成文件么，我用原论文的文件生产出的标签很不理想。万分感谢

[aoyanl]

Hi, do you got any problems in visualizing the masked labels? my visualized labels: visualized labels in this paper: It seems that my visualized results are over-masked compared with the picture in this paper.

Hi, Could you use it normally? I had the same problem as you in #10.

feel sorry to disturb you. I confused by my visualized labels. Why is the label I generated upside down compared to yours?

[Aurelien-VB]

I think that I found the cause of black areas: the code uses fillConvexPoly. By replacing cv2.fillConvexPoly(mask, polygon, value) by cv2.fillPoly(mask, [polygon], value) it fixed the issue in my case. Since fillPoly can draw multiple polygons, an optimization would be to draw them all at once.