How lsdfd/metadeepth works
MetaDeepth is a research prototype for information-optimal metasurface depth imaging, not a finished product. Its documented workflow configures an optical system, simulates point-source PSFs, evaluates depth information, renders encoded images from RGB-D samples, and scaffolds neural depth reconstruction. The graph preserves these stages and the optical, Fisher, and model internals. Relationships are based on README statements and SOURCE INDEX references; unsampled behavior is not inferred.
Main entry points: 04_depth_learning.py Depth train/eval · psf.py Point-source PSF · 01_psf_sanity_check.py PSF sanity · 02_psf_fisher_experiment.py PSF Fisher run
Components 19
Experiment stages
- System sweep
scripts/00_print_system_sweep.py - PSF sanity
scripts/01_psf_sanity_check.py - PSF Fisher run
scripts/02_psf_fisher_experiment.py - Image Fisher run
scripts/03_image_forward_fisher_experiment.py - Depth train/eval
scripts/04_depth_learning.py - Experiment outputs
outputs
Optical simulation
- System parameters
src/utils/system_config.py - YAML config
src/utils/config.py - Pupil and phase
src/optics/pupil.py - Field propagation
src/optics/propagation.py - Point-source PSF
src/optics/psf.py - PSF Fisher analysis
src/fisher/psf_fisher.py
Image forward
- Hypersim pilot
data/raw/hypersim_pilot_100 - RGB-D forward model
src/simulation/rgbd_forward.py
Depth learning
- Depth datasets
src/models/depth_data.py - Depth network
src/models/simple_unet.py - Depth losses
src/models/depth_losses.py - Depth metrics
src/models/depth_metrics.py
Other components
- Researcher
Connections 28
- Researcher → System sweep: runs
README.md - Researcher → PSF sanity: runs
README.md - Researcher → PSF Fisher run: runs
README.md - Researcher → Image Fisher run: runs
README.md - Researcher → Depth train/eval: runs
README.md - System sweep → System parameters: reads parameters
scripts/00_print_system_sweep.py - System parameters → YAML config: loads YAML
src/utils/system_config.py - PSF sanity → System parameters: reads parameters
scripts/01_psf_sanity_check.py - PSF sanity → Pupil and phase: uses phase
scripts/01_psf_sanity_check.py - PSF sanity → Point-source PSF: computes PSF
scripts/01_psf_sanity_check.py - PSF Fisher run → PSF Fisher analysis: evaluates information
scripts/02_psf_fisher_experiment.py - PSF Fisher run → Point-source PSF: computes PSF
scripts/02_psf_fisher_experiment.py - PSF Fisher run → Pupil and phase: uses phase
scripts/02_psf_fisher_experiment.py - Point-source PSF → Pupil and phase: builds field
src/optics/psf.py - Point-source PSF → Field propagation: propagates field
src/optics/psf.py - PSF Fisher analysis → Point-source PSF: evaluates PSF
src/fisher/psf_fisher.py - Image Fisher run → RGB-D forward model: renders and evaluates
scripts/03_image_forward_fisher_experiment.py - Image Fisher run → Pupil and phase: uses phase
scripts/03_image_forward_fisher_experiment.py - Image Fisher run → Hypersim pilot: loads sample
README.md - RGB-D forward model → Point-source PSF: renders with PSF
src/simulation/rgbd_forward.py - Depth train/eval → Depth datasets: builds datasets
scripts/04_depth_learning.py - Depth train/eval → Depth network: trains model
scripts/04_depth_learning.py - Depth train/eval → Depth losses: computes loss
scripts/04_depth_learning.py - Depth train/eval → Depth metrics: evaluates predictions
scripts/04_depth_learning.py - PSF sanity → Experiment outputs: writes summaries
README.md - PSF Fisher run → Experiment outputs: writes Fisher results
README.md - Image Fisher run → Experiment outputs: writes image results
README.md - Depth train/eval → Experiment outputs: writes metrics
README.md