search

InhwanBae / ETH-UCY-Trajectory-Visualizer

MIT License
11 stars 0 forks source link

why univ like this? #3

Open wza13 opened 1 week ago

wza13 commented 1 week ago

traj_1_10

InhwanBae commented 1 week ago

Hello @wza13,

The issue where the video and person do not align appears when visualizing Student001 in the UNIV scene. Since only the video for Student003 is provided, the images corresponding to Student003 should be displayed correctly.

wza13 commented 1 week ago

Thank you for your reply. I am using the code you provided and the video was also provided by you in Univ, but the above situation occurred during the process.

InhwanBae commented 1 week ago

Hi @wza13,

That was intentional. As you can see from the Original UNIV test split, there are both student001 and student003. However, the dataset only provides video data for student003, so student001 is inevitably visualized in the video of student003. That's why It visualizes like yours. if you remove the student001 file only for visualization, it will work as well as you want.

wza13 commented 1 week ago

Thank you for your patient answer. I wonder if you still have the H.txt and video.avi corresponding to the sdd data. If you can provide it, I will be grateful.

InhwanBae commented 1 week ago

The SDD dataset uses a pixel coordinate system, so you can use the homography matrix H = np.eye(3) #[[1, 0, 0], [0, 1, 0], [0, 0, 1]]. Since the original SDD dataset includes video files, you can organize the data similarly to ETH/UCY by creating separate folders named after each video and placing the corresponding video and annotation files in them. I also provide the preprocessed SDD data, which is similar to ETH/UCY.

wza13 commented 1 week ago

Thank you so much for taking the time to answer my question. Your explanation was very helpful and has clarified a lot.