search

abduallahmohamed / Social-STGCNN

Code for "Social-STGCNN: A Social Spatio-Temporal Graph Convolutional Neural Network for Human Trajectory Prediction" CVPR 2020
MIT License
483 stars 141 forks source link

Question on the dataset #33

Closed JacobYuan7 closed 3 years ago

JacobYuan7 commented 3 years ago

Hello author, I am new to trajectory prediction so I get an easy question... Are ETH and UCY dataset containing any images? I don't find any but notice that there is a background image in the main paper (i.e. in Figure 4). Where does it come from?

abduallahmohamed commented 3 years ago

Hi, The ETH images can be found here: https://data.vision.ee.ethz.ch/cvl/aem/ewap_dataset_full.tgz , this is the main website https://icu.ee.ethz.ch/research/datsets.html

The UCY images: For some reasons i can't find a source for the images, https://graphics.cs.ucy.ac.cy/research/downloads/crowd-data it's kinda down. Thus we used a fixed background to illustrate the results. More details can be found here: http://trajnet.stanford.edu/data.php?n=1

tom728 commented 3 years ago

What does the number of the txt file in datasets mean? How to generate visual results? Thank you

abduallahmohamed commented 3 years ago

@hxy-tech
The text is: For visual results: https://github.com/abduallahmohamed/Social-STGCNN/issues/9

QiyuLuo commented 3 years ago

@hxy-tech The text is: For visual results: #9 Excuse me. I am also new to trajectory prediction, I saw that ETH has such a website, https://data.vision.ee.ethz.ch/cvl/aess/dataset/ . I don't know if it is the same, but its annotation is the target bbox, but your datasets is x,y, then what does x,y stand for, how did you get it ?

abduallahmohamed commented 3 years ago

x,y are the location of ped in the real world in meters. On Thu, Jul 22, 2021 at 4:07 AM 前尘忆梦 @.***> wrote:

@hxy-tech https://github.com/hxy-tech The text is: For visual results: #9 https://github.com/abduallahmohamed/Social-STGCNN/issues/9 Excuse me. I am also new to trajectory prediction, I saw that ETH has such a website, https://data.vision.ee.ethz.ch/cvl/aess/dataset/ http://url . I don't know if it is the same, but its annotation is the target bbox, but your datasets is x,y, then what does x,y stand for, how did you get it ?

— You are receiving this because you modified the open/close state.

Reply to this email directly, view it on GitHub https://github.com/abduallahmohamed/Social-STGCNN/issues/33#issuecomment-884762437, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFIVLG3TBI5IAVM3NCQ4MKTTY7NTHANCNFSM4RWV7B7A .

--

Abduallah Mohamed abduallahmohamed.com

QiyuLuo commented 3 years ago

@abduallahmohamed Sorry, I mainly work in computer vision, so I don't know how the coordinates are calculated in meters. In computer vision, targets in an image are usually marked with bboxes (for example, upper left and lower right coordinates). I only have the coordinates of the target in the image (the upper left corner and the lower right corner are marked). Can I calculate the center point of the target Bbox (whether normalization is required) by the coordinates of these two points instead of the coordinates in meters in your data set? Or how to get the coordinates of the target in meters？ Thank you for your answer.

abduallahmohamed commented 3 years ago

The x,y are in real world coordinates; they are transformed from image domain to real-world domain by means of a homography matrix. They are labeled as points not a bounding box as this problem of trajectory estimation doesn't need the bounding box information. Also, the ETH, UCY are very old datasets so you wouldn't find bbox information there. Yet, you can guess the bbox by avg human size but wouldn't be accurate. I recall there is information about the homography matrix in the one the previously closed issues.

Please let me know if this clears or it doesn't answer your question.

Abduallah Mohamed abduallahmohamed.com

On Thu, Jul 22, 2021 at 8:09 PM 前尘忆梦 @.***> wrote:

@abduallahmohamed https://github.com/abduallahmohamed Sorry, I mainly work in computer vision, so I don't know how the coordinates are calculated in meters. In computer vision, targets in an image are usually marked with bboxes (for example, upper left and lower right coordinates). I only have the coordinates of the target in the image (the upper left corner and the lower right corner are marked). Can I calculate the center point of the target Bbox (whether normalization is required) by the coordinates of these two points instead of the coordinates in meters in your data set? Or how to get the coordinates of the target in meters？ Thank you for your answer.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/abduallahmohamed/Social-STGCNN/issues/33#issuecomment-885335624, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFIVLG6IQCWELMVY4O3YMPLTZC6K3ANCNFSM4RWV7B7A .