paranke commited on
Commit
6049be1
·
verified ·
1 Parent(s): c05d896

Document complete prepared corpus and standalone reproduction workflow

Browse files
Files changed (1) hide show
  1. README.md +8 -0
README.md CHANGED
@@ -50,3 +50,11 @@ terms; source URLs and attribution are preserved in the original manifests.
50
  ## Submission inference data (excluded from training)
51
 
52
  `submission/PHerc0211-3305-r002-c01/` contains the native CT planes 07–23, original saved Expanded9 probabilities and coverage, the pre-normalization geometry source, and the fixed 1cm scale-bar raster. This PHerc0211 region is inference-only; it is absent from the training manifest and loader. The final native tifxyz and final images live only in the linked Drive submission. Code lives only in GitHub. `validation-target.npz`, `validation-reference-label.png`, and `validation-source.json` support the historical validation result.
 
 
 
 
 
 
 
 
 
50
  ## Submission inference data (excluded from training)
51
 
52
  `submission/PHerc0211-3305-r002-c01/` contains the native CT planes 07–23, original saved Expanded9 probabilities and coverage, the pre-normalization geometry source, and the fixed 1cm scale-bar raster. This PHerc0211 region is inference-only; it is absent from the training manifest and loader. The final native tifxyz and final images live only in the linked Drive submission. Code lives only in GitHub. `validation-target.npz`, `validation-reference-label.png`, and `validation-source.json` support the historical validation result.
53
+
54
+ ## Reproduction entry point and storage
55
+
56
+ Follow the [numbered GitHub guide](https://github.com/priyankaranke/vesuvius-expanded9-training#readme) for download → preparation → scratch training → weight export → inference → exact viewer export. All 113 prepared representations (112 training, one validation) and all 130,645 CT tiles are already here; no training sample needs a new CT render. The corpus is about 30.6 GB: 29.24 GB compressed CT plus approximately 1.4 GB labels, coordinates and source metadata. Allow about 70 GB locally when retaining the download, restored tiles and checkpoints.
57
+
58
+ The default preparation step restores the original compressed tile files from the byte packs; it does not render meshes or resample CT. Once downloaded, training needs no original workstation, historical cache, source-site login or W&B service. The historical source URLs remain provenance; the full input corpus is supplied here. Human-approved labels are preserved.
59
+
60
+ Normal authorized Hub file downloads are sufficient. HF Pro and the paid private-dataset viewer are not required. Read access must cover this dataset and the model repository; private URLs alone do not grant permission. Code and native rendering helpers are supplied in the one linked GitHub repository. The final mesh and PNGs remain on Drive.