xjxu21 commited on
Commit
2aa9dbf
·
verified ·
1 Parent(s): a2f526a

docs: record the per-scene render provenance difference

Browse files
Files changed (1) hide show
  1. README.md +17 -0
README.md CHANGED
@@ -69,6 +69,23 @@ Clips are 11 s to 8.5 min of continuous motion — no cuts, no teleports.
69
  * **Panoramic frames are equirectangular.** A rectangular crop is *not* a perspective view —
70
  reproject before comparing against the first-person clip.
71
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
72
  ## Tools
73
 
74
  ```bash
 
69
  * **Panoramic frames are equirectangular.** A rectangular crop is *not* a perspective view —
70
  reproject before comparing against the first-person clip.
71
 
72
+ ## Render provenance
73
+
74
+ `description.json` records the exact render path per clip, and it is **not identical across
75
+ scenes**. Five scenes (med_village, venice, office, apartment, art_nouveau) were rendered in
76
+ two rounds — `MPPC_RGBOnly` then `MPPC_DepthPlus` — and stitched with a Lanczos kernel, so
77
+ `render.mppc` is an object with `rgb`/`depth` keys. **paris** was re-rendered later through a
78
+ corrected single-pass path (`MPPC_VelocityDepthPlus` with the extra pass dropped, linear-HDR
79
+ output, fixed exposure, linear stitch kernel), so its `render.mppc` is a plain string and its
80
+ `render.cube` additionally carries `fixed_ev` and `tm_comp`.
81
+
82
+ The reason for the paris re-render: a Lanczos resampling kernel has negative lobes, which on
83
+ high-contrast sky/architecture boundaries interpolate linear HDR below zero; clamping to zero
84
+ left a one-pixel pure-black rim along building silhouettes. Switching to a linear kernel took
85
+ sky edges with undershoot from 15.5% to 3.4% — below the 6.1% baseline of the cube faces
86
+ themselves. Per-frame auto-exposure was replaced with a fixed value at the same time, so
87
+ brightness no longer drifts with what happens to be in view.
88
+
89
  ## Tools
90
 
91
  ```bash