VisDrones object detection dataset how to generate density map and next step crop ？

[zxq309]

VisDrone Crowd Counting dataset Ground Truth is in xml format, do you means to convert it to mat format and then send it to MCNN network to generate density map？ how to generate density map for VisDrones object detection dataset, its Ground Truth is different，please help me，thanks.

[Cli98]

VisDrone Crowd Counting dataset Ground Truth is in xml format, do you means to convert it to mat format and then send it to MCNN network to generate density map？ how to generate density map for VisDrones object detection dataset, its Ground Truth is different，please help me，thanks.

Hi Zhang,

We only use 2018 DET dataset. You need to 1. extract center of all objects and then 2. run a Gaussian kernel on your map to get result.

Actually, what you asked have been provided. Check function gaussian_filter_density in https://github.com/Cli98/DMNet/blob/master/image%20cropping/Generate_density_map_official.py. Run this file to get density map.

Cheers,

[zxq309]

VisDrone Crowd Counting dataset Ground Truth is in xml format, do you means to convert it to mat format and then send it to MCNN network to generate density map？ how to generate density map for VisDrones object detection dataset, its Ground Truth is different，please help me，thanks.

Hi Zhang,

We only use 2018 DET dataset. You need to 1. extract center of all objects and then 2. run a Gaussian kernel on your map to get result.

Actually, what you asked have been provided. Check function gaussian_filter_density in https://github.com/Cli98/DMNet/blob/master/image%20cropping/Generate_density_map_official.py. Run this file to get density map.

Cheers,

Thanks for your reply! I run "Generate_density_map_official.py" and "density_slide_window_official.py" successfully on your answers. However, I have another question when I run "fusion_detection_result_official.py" can you help me?

[Cli98]

VisDrone Crowd Counting dataset Ground Truth is in xml format, do you means to convert it to mat format and then send it to MCNN network to generate density map？ how to generate density map for VisDrones object detection dataset, its Ground Truth is different，please help me，thanks.

Hi Zhang, We only use 2018 DET dataset. You need to 1. extract center of all objects and then 2. run a Gaussian kernel on your map to get result. Actually, what you asked have been provided. Check function gaussian_filter_density in https://github.com/Cli98/DMNet/blob/master/image%20cropping/Generate_density_map_official.py. Run this file to get density map. Cheers,

Thanks for your reply! I run "Generate_density_map_official.py" and "density_slide_window_official.py" successfully on your answers. However, I have another question when I run "fusion_detection_result_official.py" can you help me?

Apologize for late reply. The result are not in expected coco format. It will help if you post format for your generated coco file. Let me know if this helps u.

[wyl-19980129]

@XiangqingZhang 你好，请问你在融合检测（fusion_detection_result_official.py）前，使用faster-rcnn分别对原图像和裁剪后的小图片进行检测时，config文件是如何配置的呢？看作者这里只提供了预训练参数，并没有提供对应于预训练参数的config文件。

[Mik030]

@XiangqingZhang 你好，请问你在融合检测（fusion_detection_result_official.py）前，使用faster-rcnn分别对原图像和裁剪后的小图片进行检测时，config文件是如何配置的呢？看作者这里只提供了预训练参数，并没有提供对应于预训练参数的config文件。

直接使用mmdet中的config文件即可。

[Cli98]

@XiangqingZhang 你好，请问你在融合检测（fusion_detection_result_official.py）前，使用faster-rcnn分别对原图像和裁剪后的小图片进行检测时，config文件是如何配置的呢？看作者这里只提供了预训练参数，并没有提供对应于预训练参数的config文件。

直接使用mmdet中的config文件即可。

Mik030说的是对的，直接使用mmdetection V2.0的配置文件就行

[Cli98]

Closed as no activity by far. Feel free to re-open for any further questions.