shift-dev
shift-dev copied to clipboard
SysCV
SHIFT Dataset DevKit - CVPR2022
I see many missing and wrong annotations in 2D/3D/InstanceSegmentation. Example in "https://dl.cv.ethz.ch/shift/discrete/images/val/front/det_2d.json" : * name : 00000220_img_front.jpg * videoName: 773a-45da * Only 3 vehicles are present in the annotations, but...
Hello, I am currently conducting research as an undergraduate student with a specific focus on the test-time adaptation for object detection problem. I have implemented some methods based on the...
Hi, thanks for generating this awesome dataset. It appears you only have optical flow for images frames (1fps). Is there a way to get the optical flow for the video...
There are these towns: ` {"01" | "02" | "03" | "04" | "05" | "06" | "07" | "10HD"} ` Do the correspond to rural, urban etc. styles? What...
I tried different versions and wondered why it didn't work. Turns out, that python
Hi,all I used the depth convertion formula（Depth (meters) = (256 * 256 * B + 256 * G + R) / (256 * 256 * 256 - 1) * 1000...
Here's my code. I am trying to read the optical flow to warp frame 2 back to frame 1. But the results appear to be wrong. ``` flo = np.load(r'E:\Dataset\SHIFT\flow\0003-17fb\00000040_flow_front.npz')...
  Why in continuous/100x training set, the file size of det_insseg_2d.json in image is the same as video (I think the size of the label file in the video...
Thank you very much for your data set. When I visualized the 3D box, it would visualize some samples blocked by the wall. Can this be processed? 
