The committed pipeline.yaml only had the pipeline: operations block, missing the parameters: section (xlims, zlims, dynamic_range, f_number, apply_lens_correction) that reconstruct.py applies separately in code. Two consequences of the missing parameters:
- dynamic_range defaulted to zea.display.to_8bit'''s built-in (-60, 0) dB instead of this dataset'''s actual [-50, 0], washing out blacks in the reconstruction.
- grid_size_x/grid_size_z were left to auto-derive with no xlims/zlims constraint, using the full native depth range instead of skipping the first 2mm dead-zone, producing a visibly different (and much taller) scan cone than the reference reconstruction.
This adds the parameters: block so [1m[38;5;36mzea[0m[0m: Using backend 'jax'
-------------------GPU settings-------------------
memory
GPU
0 45458
1 45458
2 37654
3 45458
4 45458
Selecting 1 GPU based on available memory.
Selected GPU 0 with Free Memory: 45458.00 MiB
Hiding GPU…
- Tidy the pipeline YAMLs (7e726adfd8bd3e8ab397e8a4fb284fe20605c151)
- Ship the two configs the reconstruct scripts need (d330799806f927124f06b01ed0617f690271799c)
- Normalize line endings to LF (cce1ebbf5636b564fb362764eec038afbc1cd7d9)
- Remove the Python helpers from twente-microbubblesim (0fde458d1f28fe3ff3d6ba1ceac2a6cf0ee16411)
- Remove the last reconstruction script (5d49e4f4dc1eccd3cf2986a45e9cb493eb6ee017)
- add cross link to github (0715e04468208c71223097ca6ac292594882f6ce)
- Drop the comment from .gitattributes (33fd6aacd7e112ea169d1ca953c475e81362ca5a)
- Point data cards at pipeline.yaml and the merged reconstruct scripts (2de063e00e78dfe4f0acd302fa4ded18740f5a56)
- Lead the tue-carotid card with a rendered bifurcation loop (1df230e2f455ed02a31177b9e80eebba53424a29)
- Match the wpi card's code block to the formatter (cbe25eb94fa1b9405d40574ff354c8194b38b095)
Co-authored-by: Tristan Stevens <tristan-deep@users.noreply.huggingface.co>
30 days ago
Simplify root README and standardize institution credits (#4)
