Jiacheng Chen

Hi, Basically we compute the optical flow of every two ajacent frames and then stack them together to form the optical flow for the whole sequence. So when the number...

I think the frames in your screenshot are good : ), after you crop the human and resize all frames into the same size, you can compute the stacked optical...

I think the two-stream model assumes the input optical flow to be from a single video sequence. It's not good to stack flows extracted from 8 videos from different angles,...

Please refer to [this file](https://github.com/woodfrog/ActionRecognition/blob/master/utils/UCF_preprocessing.py) for details. And also see [this file](https://github.com/woodfrog/ActionRecognition/blob/master/utils/OF_utils.py) for how to generate optical flow for videos.

Hi, I believe that's because the processing is not finished as expected so that the necessary files are missing. Please check [this line](https://github.com/woodfrog/ActionRecognition/blob/master/utils/UCF_utils.py#L37) where the error is raised.

Hi, you could try to adapt the data preprocessing script according to your own data format, and you can simply remove all code related to the annotation processing for running...

@volodymyrkepsha Hi, your first density map looks all right, and I believe the pretrained Mask-RCNN and corner/edge models should be able to produce okay results for that one. There might...

Hi, you need to fill the form here [http://realsee.com/website/open/en](http://realsee.com/website/open/en), then Beike's system will send the data download links to your email.

Hi, Thank you for your interest. We used the [evaluation code of FloorNet](https://github.com/art-programmer/FloorNet/blob/master/evaluate.py) to get the numbers in the paper (with some simple modification for our output format).

Uh, sorry for the confusion. The file utils.py was directly copied from FloorNet's codebase... FloorNet provides the functions for computing recall/precision for walls(edges) and rooms, these functions can be adapted...