sutdcv / uav-human Goto Github PK

According to annotation P070, the PersonID is a unique identification for each person. However, it is unclear why the code uavhumanreidtools/datasets/uavhuman.py returns the index of the unique identifier when relabel is True, which can be confusing.

Hi,

the dataset does not contain a train/test split for the poses. Is there missing a file defining it? I also do not have the train test directories as shown in uavhumanposetools. I only have a Skeleton folder with all the .txt files.

Thank you!

Hi,
I downloaded files from your UAVHuman/ActionRecognition/RGBVideos directory but their format is weird. How to convert these files to a proper video files? Also where can I find labels to them and RGB-D videos?

Hi, thank you for your great work! Will RGB videos be released for action recognition task? Currently I only see the fish-eye videos and skeleton in action recognition folder. Thanks!

Hello, I found that the most of the angles are horizontal in the images of person reid in the dataset.
It doesn't look like a drone camera shot.

It is something wrone? Or I download yhe wrong data?

Hi, it seems that the hat and backpack file name cannot match with the label set provided in the README.md. For example:

P015S00G10B50H50UC022000LC022000A001R0_08261237_055_bbx.jpg

consists of "B50" and "H50" while the hat and backpack label set is only up to 4. May I ask what should I do to fix this?

Thanks for your clarification!

I downloaded the human pose dataset from the Google drive you gave me, and then divided the dataset according to the operation of uavhuman pose tools, but there was a problem: filenotfounderror: [errno 2] no such file or directory: '/ poseestimation / train_ data. npy'。 Is there a problem with the dataset I downloaded, which only contains frames and annotations.

test dataset file only have xxxx.txt. miss classes

When the attribute data will be provided in the Google Drive?

Thank you for your great work!

Do you have the text of action labels, for example, class A001 denotes jumping. Thank you!

Hi,

Will you release the code of st-gcn, shift-gcn, and other methods evaluated on this dataset?
Since it would be easier to use this dataset by acknowledging how to modify the code when using this new dataset.

Good day to you
Reid and par datasets contain same labels per an image, i guess, but have different the number of images in your paper.
(PAR : 22,263 images, ReID : 41,290 images)
I wonder why there is a difference between the number of both datasets.
Thank you for sharing your great work!

Hi, thanks for your work!
I have downloaded the skeleton-based action recognition dataset, but I can't find instructions about the txt file format. So what is the detailed format of the content of each txt file in the skeleton dataset?

Is there some IR / RGB image for using?

Thank you for the great work! Is the depth data released? I couldn't find it in the Google Drive. Thanks!

waiting for your open source dataset.

In this hub, you only provided the partitioning method for train, but the partitioning of query and gallery has not been updated. May I ask how the pedestrian ID and camera ID of query and gallery are distinguished in this dataset?thanks for your reply.