capture.md

How to capture recoils

1. (one time) install python 3, pipenv. Run pipenv install in ./processing dir. You will also need to record in-game footage - NVIDIA overlay or OBS worked fine for me.

   Update Apex Legends starup options to include +cl_showpos 1 .

2. Go to fire range and take a weapon with maximum mag and no other attachments and a sniper rifle with x4-8. Start recoding, fire the whole magazine then point to the most distant points through sniper scope, capture the whole pattern too. I have used avidemux player to navigate between frames. Here is an example recording:

   It's important not to move mouse while shooting. One can either lift a mouse from a pad or map mouse buttons to keyboard if possible.

   Save pattern image and crop it:

   

3. Open ./processing/recoils.ipynb jupyter notebook and enter angle coordinates of two distant points into np.round(points(...), 2) cell. That will give you a distance in pixels between points. Open https://apexlegendsrecoils.net/editor, load an image of the pattern, enter name of the weapon (see ./client/specs.json for possible values). Use left mouse click to add and move individual points, right button to remove them and middle button to mark "anchors". There should be exactly two anchors at marks that we just measured. "auto targets" might also be useful as it converts detected areas to marks - play with settings to understand how it works. Then select starting mark and hover over points to create a path. Normally y-coordinate in the result should be negative as it offsetts gun upward movement.

   Copy the resulting json string back to jupyter notebook rc = [] . Running next cell will produce arrays for X and Y coordinates that now can be put into ./client/specs.json. This coordinates correspond to mouse positons for sensitivity 1.

   

   As there is some randomness of the pattern - repeat 2-4 steps a few times - there is a "error estimation" cell that might help to understand if results have converged enough.