The demo code for DM Count is included in crowd.py , which is modified version of the original repo implementation. The procedure of normal data load has been replaced by loading from epoch wise json files as shown in link
The dataset-wise epoch json files can be found at : zenodo link
In this folder a demo of how the images should be sampled from bins is provided as a gif below.
| Round Robin Sampling | Random Sampling |
|---|---|
 |
 |