Add camera calibration steps to pre-experiment checklist

[glopesdev]

Related to #168, #107 and #78, we need to develop a reproducible pre-experiment checklist for quality control of camera image. This should allow us to control for illumination, focus, gain, filter, lens, and sensor changes to minimize unexpected divergence in quality for a known calibration target.

As a first protocol, we will use the laminated rigid checkerboard already used for intrinsic camera calibration (see Calibration Targets). For this issue to be resolved, the README should be updated to include camera quality checks. For this, we need to resolve the following:

• [ ] Decide standard positions for calibration targets to maximize arena coverage
• [ ] Decide calibration data format:
  • [ ] Illumination parameters:
  • [ ] Type of light source (e.g. LED, spotlight)
  • [ ] Make and model
  • [ ] Power supply parameters
  • [ ] Number of bars / spotlights
  • [ ] Spatial distribution (top, ring-light, scattered pointing up, etc)
  • [ ] Lens parameters (make and model)
  • [ ] For each camera:
  • [ ] Make and model
  • [ ] Sensor model
  • [ ] Acquisition parameters (gain, binning, exposure, etc)
  • [ ] Rough spatial location (overview, side, quadrant, etc, also see #154)
  • [ ] Segmentation mask (optional)
  • [ ] IR Filter parameters (make and model, spectral response, efficiency, etc)
• [ ] Decide version control mechanism (dataset vs aeon_experiments git repo)
• [ ] Update calibration workflow to record both pre- and post-calibration screenshots
• [ ] Develop set of reproducible analysis scripts for benchmarking image quality
• [ ] Checks for streaming from all camera devices

[bruno-f-cruz]

• "Decide standard positions for calibration targets to maximize arena coverage"

Wouldn't the use of QR/Aruco Markers be better suited for this? Using a checkboard will make the procedure "acute" and dependent on a human to put the markers in the correct spot. If we have the makers permanently in place it would be much easier. It would also be easier to infer scale, focus etc.. in multiple places in the arena.

[glopesdev]

@bruno-f-cruz we did discuss including Aruco markers in each hexagon tile of the arena, and I agree it would help a lot also with position of patches and other objects in the experiment. However, for calibration we probably still need to have "flying" checkerboards close to the camera so that we can correctly infer lens distortion, etc.

We could replace the traditional checkerboard with a ChArUco board, but we would have to reprint the laminated metal board we have now (which is really accurate). We could definitely do it, but it seems like the main advantage of ChArUco over a normal board is to allow for partially occluded views of the chessboard, which might not be a deal breaker in our case (since we are in control we can probably find some reasonable configurations that work) so I'm not entirely convinced we have to do it to get some good first results.