dar96io commited on
Commit
2d5d49b
·
verified ·
1 Parent(s): f48664a

Upload folder using huggingface_hub

Browse files
README.md CHANGED
@@ -1,3 +1,34 @@
1
  ---
2
- license: mit
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ language: en
3
+ library_name: pytorch
4
+ pipeline_tag: image-to-image
5
+ tags:
6
+ - diffusion
7
+ - image-restoration
8
+ - remote-sensing
9
+ - underwater
10
  ---
11
+ # KAUSTics-Net: selected result checkpoints
12
+
13
+ Pixel-space conditional velocity-diffusion model for through-water UAV restoration. This model has no latent autoencoder. Use the accompanying KAUSTics-Net GitHub code; these files are not Diffusers pipeline checkpoints.
14
+
15
+ ## Contents
16
+
17
+ | Directory | Seed checkpoints | Initialization / selected epoch |
18
+ |---|---|---|
19
+ | `kaustics/` | `seed1.pt`, `seed2.pt`, `seed3.pt` | Scratch KAUSTics training; epochs975/925/925 |
20
+ | `sea_undistort/` | `seed1.pt`, `seed2.pt`, `seed3.pt` | Corresponding KAUSTics seed, sampler-aware fine-tuning; epochs95/100/100 |
21
+
22
+ Only the six selected model weights are included. No optimizer states, intermediate epochs, pilots, ablations or competitor weights. `selected.json` records SHA256 checksums, original archival checkpoint checksums and unchanged tensor digests. The lightweight release exports CPU tensors plus primitive metadata; load with `torch.load(..., map_location='cpu', weights_only=True)`.
23
+
24
+ ## Inference
25
+
26
+ Use `infer.py --checkpoint checkpoints/kaustics/seed1.pt --input image.png --output predictions/example` in the code release. Input/output256×256 RGB; input normalization[-1,1];1000-step cosine schedule;25-step DDIM at strength0.95 and classifier-free guidance2.0. Preserve image filenames for deterministic per-image noise.
27
+
28
+ KAUSTics training uses source-disjoint26/3/3 UAV source IDs. Sea-Undistort fine-tuning uses840 all-effects training scenes and240 all-effects validation scenes; all three120-scene test variants are scored against original GT. See the code release’s `docs/PROTOCOL.md` for loss weights, selection, stopping rules, benchmark exposure and portability limits.
29
+
30
+ These weights target the paper’s dataset/task. Restoration quality and geometric accuracy on unrelated real UAV scenes are not guaranteed. Sea-Undistort adaptation shows a fidelity/perceptual trade-off; it is not a claim of universally improved image quality.
31
+
32
+ Authors: Dario Scilla, Omar A. Lopez, Kasper Johansen, Brian O. Nieuwenhuis and Matthew F. McCabe. Paper: *KAUSTics-Net: Diffusion-Based Through-Water UAV Restoration with a Semi-Synthetic Dataset* (revision manuscript).
33
+
34
+ Licensing for the authors’ model weights will be specified by the authors before publication; this prepared card does not assign a license. Dataset terms remain separate.
SHA256SUMS.txt ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ 5dda11d4d552e14cdd7bf50c8dc58fb91322906f455b2339abf6c567d71ebc85 README.md
2
+ b75786cc04deb435f600169e40bbff0bc2b3780661ec1f20d7f8a3bba9901b74 kaustics/seed1.pt
3
+ 2d75d39650e182c08204f4f46cd385d2fee55d64b072feafa5426aca4c63c22b kaustics/seed2.pt
4
+ b7c38902ea7571e9744e78ec264d8d7db19418e9b64f85434e527b9fc93a6213 kaustics/seed3.pt
5
+ 66a23614353efe59649ad973103a93240cd355d94ee673bfd3ddf131f5e4199b sea_undistort/seed1.pt
6
+ fff8b6c8a454af176f19bd770f36bbe2ee2af26ad59fda05568e753337bdd0a1 sea_undistort/seed2.pt
7
+ dd28190af24b4a195cf0e5a1848778ec27020b7abb441292ba3853e236c67cff sea_undistort/seed3.pt
8
+ 5be3da7edf108a9ac4e51baaef722131bc4264d97b123ac542d6229b4106b70e selected.json
kaustics/seed1.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b75786cc04deb435f600169e40bbff0bc2b3780661ec1f20d7f8a3bba9901b74
3
+ size 26466546
kaustics/seed2.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2d75d39650e182c08204f4f46cd385d2fee55d64b072feafa5426aca4c63c22b
3
+ size 26466546
kaustics/seed3.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b7c38902ea7571e9744e78ec264d8d7db19418e9b64f85434e527b9fc93a6213
3
+ size 26466546
sea_undistort/seed1.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:66a23614353efe59649ad973103a93240cd355d94ee673bfd3ddf131f5e4199b
3
+ size 26461618
sea_undistort/seed2.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fff8b6c8a454af176f19bd770f36bbe2ee2af26ad59fda05568e753337bdd0a1
3
+ size 26461618
sea_undistort/seed3.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd28190af24b4a195cf0e5a1848778ec27020b7abb441292ba3853e236c67cff
3
+ size 26461618
selected.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "checkpoints": [
3
+ {
4
+ "path": "kaustics/seed1.pt",
5
+ "seed": 1,
6
+ "family": "kaustics",
7
+ "epoch": 975,
8
+ "sha256": "b75786cc04deb435f600169e40bbff0bc2b3780661ec1f20d7f8a3bba9901b74",
9
+ "model_sha256": "dee37dc07bf6d51782c419c7ecc4c2cfdc1242089b2307409045c4dbe8e9f5ec",
10
+ "original_checkpoint_sha256": "eac94b3d7c5b364c5e212f3f748d2e6a94d392a0f943d92b3fe0ee9e09d70edb",
11
+ "bytes": 26466546
12
+ },
13
+ {
14
+ "path": "kaustics/seed2.pt",
15
+ "seed": 2,
16
+ "family": "kaustics",
17
+ "epoch": 925,
18
+ "sha256": "2d75d39650e182c08204f4f46cd385d2fee55d64b072feafa5426aca4c63c22b",
19
+ "model_sha256": "da2df168563c79bfbcc131a1f2bfed6868e5596850f26aaee645d4360802d29e",
20
+ "original_checkpoint_sha256": "4c9b3daa082c09ceb8630abd6495014d2e1746d42010868893501fe03ed18776",
21
+ "bytes": 26466546
22
+ },
23
+ {
24
+ "path": "kaustics/seed3.pt",
25
+ "seed": 3,
26
+ "family": "kaustics",
27
+ "epoch": 925,
28
+ "sha256": "b7c38902ea7571e9744e78ec264d8d7db19418e9b64f85434e527b9fc93a6213",
29
+ "model_sha256": "6eb0650764e56be3d56f815924f70d0c78191e8a46964d9bf1dcd6d076ee1df9",
30
+ "original_checkpoint_sha256": "7d769a2e9b00eec16da90d0de8cf1f48d501617253e38351b73d2341e8293314",
31
+ "bytes": 26466546
32
+ },
33
+ {
34
+ "path": "sea_undistort/seed1.pt",
35
+ "seed": 1,
36
+ "family": "sea_undistort",
37
+ "epoch": 95,
38
+ "sha256": "66a23614353efe59649ad973103a93240cd355d94ee673bfd3ddf131f5e4199b",
39
+ "model_sha256": "df81f57bcf7f4d13f04db3e43fff02ec822889d709235a1e58ae09234be953e8",
40
+ "original_checkpoint_sha256": "c6764d99586862e6acbbd7a0c92aece25db16172dd8ce160cfa216482150c476",
41
+ "bytes": 26461618
42
+ },
43
+ {
44
+ "path": "sea_undistort/seed2.pt",
45
+ "seed": 2,
46
+ "family": "sea_undistort",
47
+ "epoch": 100,
48
+ "sha256": "fff8b6c8a454af176f19bd770f36bbe2ee2af26ad59fda05568e753337bdd0a1",
49
+ "model_sha256": "e1201b07c9b56024ce9462f1f9c1d7311b42dcb3fb92f66539cf5b23581db3e2",
50
+ "original_checkpoint_sha256": "e35a7e4517d235cf46e1e682b5b66d5f7c17a3763af00d8fc70a43587c49cdf4",
51
+ "bytes": 26461618
52
+ },
53
+ {
54
+ "path": "sea_undistort/seed3.pt",
55
+ "seed": 3,
56
+ "family": "sea_undistort",
57
+ "epoch": 100,
58
+ "sha256": "dd28190af24b4a195cf0e5a1848778ec27020b7abb441292ba3853e236c67cff",
59
+ "model_sha256": "758cf063e2326827d5f471ce50fd4f71cfd5c362142c56bcdae9db4e08dd7901",
60
+ "original_checkpoint_sha256": "d8f6fb99b04ec8b36b5a51df956ba3c7cc544e5a6e060276d8c481c72ff9a376",
61
+ "bytes": 26461618
62
+ }
63
+ ]
64
+ }