From e9398de9c1c6084626cffb5033dca99d41fc20cc Mon Sep 17 00:00:00 2001 From: Duncan Tourolle Date: Sun, 20 Sep 2026 05:14:50 +0200 Subject: [PATCH] Ship the fine-tuned border filler: MI-GAN 512 trained on projection-shaped voids from the maintainer's library --- docs/panorama.md | 23 ++++++++++++++++++++++- models/inpaint/migan-512.onnx | 4 ++-- 2 files changed, 24 insertions(+), 3 deletions(-) diff --git a/docs/panorama.md b/docs/panorama.md index 4ee555c..904efcc 100644 --- a/docs/panorama.md +++ b/docs/panorama.md @@ -522,4 +522,25 @@ coarse 1 (the coarse pass seeds nothing the model is allowed to see), band above zero. The model file is a drop-in (`models/inpaint/migan-512.onnx`, same six operators, same tensors) and the engine loads it unchanged. -**Measured.** PLACEHOLDER +**Measured, 240 held-out tiles with projection-shaped voids (PSNR in the +hole, dB / LPIPS on the composite), stock → shipped (step 3 607):** edge +16.9 → 18.5 / 0.121 → 0.136; corner 14.6 → 16.3 / 0.183 → 0.205; interior +18.2 → 19.3 / 0.051 → 0.056. Read both columns: the fine-tune gains ~2 dB +on edges and corners because it stops inventing objects, and *loses* on +LPIPS because what it paints in a deep void is smoother than the stock +model's confident wrong texture — LPIPS rewards texture, right or wrong. +On the fixture's dump at half resolution (the merge's working size) the +sky corners are sky, with no structure and a faint tone step at worst; +the ground bands carry a fine texture at the right tone, softer than the +real scree above them. The stock model's top-left corner on the same +dump is a glowing invented structure. The training's own record — what +each loss weighting did, and the two runs abandoned (blur under L1 in +the hole; a brick pattern under a strong adversarial term against a +discriminator that had not learned) — is `runs/` in `darkroom-infill`. + +**What is still wrong.** The ground fill is softer than its context — +texture, not structure, is what a night on a laptop GPU could not finish. +The levers, in order: a discriminator that learns (a pretrained one — +MI-GAN's own from the unfused checkpoint — instead of a PatchGAN from +scratch), feature matching, and more steps at 512. FR-MRG-4's +*experimental* stays. diff --git a/models/inpaint/migan-512.onnx b/models/inpaint/migan-512.onnx index 2331b79..41211ef 100644 --- a/models/inpaint/migan-512.onnx +++ b/models/inpaint/migan-512.onnx @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:42dede5a5fa86891443cacf6c976cfa81641f5870bd69b71ca5d2f907c1c73d8 -size 28009057 +oid sha256:c44dfe0740662f875cfd59ac119ebb39828ffc6a5e3f8fdbed02e356d08a4c52 +size 29604484