Heliotrope
UCL Computational Photography and Capture - Heliotrope Lab
Folder structure
- Place optical flow data in
data/{dataset_name}_flows.mat
- For example
data/gabe_flows.mat
ordata/trump_flows.mat
- You can download the Trump optical flow data from Dropbox (1.83 GB)
- For example
- Image sequence should be in the folder
data/{dataset_name}
- E.g.
data/gabe/
ordata/trump/
- E.g.
Usage
- Run
heliotrope
- Edit the variable
dataset
inheliotrope.m
if you wish to change the dataset
- Edit the variable
- Enter your desired starting image, or use the default recommended one by just pressing
Enter
- Draw as many line segments as you want
- Backspace to delete previous point
- Double-click your final point to end selection
- Wait a minute...
TIP: Use the tip of the nose as starting point to deliver best results.
Output
3 video figures will appear:
- Slow motion interpolation of trajectory-based cost
- No interpolation - Without trajectory cost
- No interpolation - With trajectory cost
Adjustable parameters
- Inside
src/best_path.m
you can editalpha
to modify the weighting of the trajectory cost.
Generate optical flow
- Run
generate_flow