-
Hi, @AlexeyAB . In caltech and citypersons dataset, there are lots of "ignore" label, which may be "a group of people” or the advertisement. So, in order to avoid the influence of these samples, how…
-
Hello,
Before anything thanks for sharing your amazing work.
I recreating the train results using the "mix" dataset and the DLA34 model.
I downloaded and followed the instructions in the rea…
-
论文中写道在coco和CityPersons上进行了实验。是否是需要将这两个数据集转换成与CrowdHuman相同的格式?是否可以提供实现格式转换的文件?
-
想请教一下作者,在数据集cityperson上训练需要做除了代码中给的数据增强之外的一些增强方法扩充网络训练的数据集数量吗?如果用cityperson的子集resonable训练网络会出现过拟合现象吗?
-
Hi there,
I check the evaluation results reported in the paper and the result from the official benchmark (https://bitbucket.org/shanshanzhang/citypersons/src/default/).
However, I find the resul…
-
Fist, thanks for your good work. I am a little confused about the lable format. For the annotations of the pedestrian dataset, the annotation has 'ignore', 'pedestrians', 'sitting people', 'others' an…
-
![image](https://user-images.githubusercontent.com/23293902/53930766-7bbd7d00-40cd-11e9-8ee0-8a84facc6be2.png)
这个结果和Repulsion loss一样都变差了,也是奇怪的很!希望作者贴出他的实验结果!
-
I believe there is some error with the splits of your data: The test.txt set contains 2299 lines, the train.txt contains 2303 lines and the eval set contains 398 lines.
But according to the origina…
-
I downloaded the BDD100K dataset and converted the labels to ms coco format using the git repo from here:
https://github.com/ucbdrive/bdd-data
I used the following command to convert the two .js…
-
Good day to you!
I have some questions about finetuning on the reid task.
1. How did you finetune the reid task in Table 3 of the paper?
2. Did you finetune separately on each dataset or jointly …