Skip to content

PixelOE 1.0.0

Latest
Latest

Choose a tag to compare

View all tags
@github-actions github-actions released this 24 Sep 20:37
v1.0.0
1d45ba0

PixelOE 1.0.0

Changes

  • chore(release): bump version to 1.0.0 (1d45ba0)
  • fix(torch): drop unused tqdm import (eae1cb8)
  • docs: document the Slang backend and backend selection (b866fff)
  • ci: release pipeline with nightlies and automatic patch releases (6f280af)
  • ci: Python 3.12-3.14 on Linux and Windows with Slang coverage (918e223)
  • build: make slangpy a platform-conditional dependency (7f0d9a8)
  • bench: run torch reference paths with backend=torch (05ce754)
  • test: pin the torch oracle and cover backend selection (dc9944e)
  • feat: automatic backend selection in pixelize() (8712429)
  • style: apply black formatting (bd72d67)
  • docs: slang vs torch benchmark README (251a914)
  • bench: paper figures and overview page (4ce77a4)
  • bench: example output exporters (059722b)
  • bench: CPU kernel ISA report (b885823)
  • bench: device comparison across CPU workers, RTX 4090 and Arc Pro B50 (ed26c32)
  • bench: scaling sweeps over size, batch and pixel size (d9330f0)
  • bench: slang vs torch config matrix runner and analysis (ddec05d)
  • test(slang): end-to-end pixelize conformance (23a2b55)
  • test(slang): quantization and dither conformance (46b3e3b)
  • test(slang): downscale conformance with exact k-centroid tie rule (fbb5b74)
  • test(slang): outline and sliding-window stats conformance (dfc7c85)
  • test(slang): color op conformance (501f197)
  • test(slang): shared fixtures and comparison helpers (089b4db)
  • feat(slang): command-line converter (3b7c07e)
  • feat(slang): pixelize pipeline matching the torch API (6520d25)
  • feat(slang): k-means quantization and dithering (9e55343)
  • feat(slang): unsharp and laplacian sharpening (4296057)
  • feat(slang): padding, interpolation and nearest-exact upscale (78e77f8)
  • feat(slang): contrast and k-centroid downscale (39318dd)
  • feat(slang): outline expansion with lattice and sliding-window stats (c365c49)
  • feat(slang): color conversion, blur and color matching (578ed8e)
  • feat(slang): segmented reductions and Lab moments (0d33433)
  • feat(slang): common shader library (a13e6b2)
  • feat(slang): backend registry and context factory (ef16784)
  • feat(slang): native CPU backend (093c4f0)
  • feat(slang): GPU device context with pooled buffers and torch interop (cfd2a10)
  • feat(slang): toolchain setup and shader/cache paths (29bd375)
  • build: add slang extra and package shader sources (efd7ecc)
  • chore: track tests in nested test directories (1574500)
  • fix: resolve GitRoll-flagged bugs and unused locals (341aa85)
  • chore(experiments): add outline mapping sweeps (f7c0ae1)
  • test: add CPU regression coverage and CI (653b814)
  • feat(ui): expose outline weight mapping controls (b1cbf80)
  • refactor(torch): formalize outline weight mapping (8fe5bbb)
  • avoid deprecated packages (c4348da)
  • formatting (d1c5e28)
  • demo: add force png, download button with filename, no upscale (8b6381d)
  • demo: add the option to quantize after the pixelate step (fc5d2b3)
  • demo: automatically adapt ratio at upload (b9b1588)
  • complete requirements (eef833e)
  • update icon (ff3b8ab)
  • update icon (9a34c09)
  • Fix wrong positional arguments (cd94a6a)
  • fix no_post_upscale (66c072e)
  • update pixel.py (210ecd2)
  • chore(publish): update workflow for node publishing (b308acf)
  • Update pyproject.toml (bc0a43f)
  • Update pyproject.toml (b94cf57)
  • Create publish.yml (2384577)
  • Update for Comfy registry (b082938)
  • Add sharpen in after OE (756a780)
  • Fix project.scripts (#9) due to cli_legacy.py rename (dd316f3) (4362aa9)
  • Update README.md (81d5feb)
  • 0.1.4 (4a676a5)
  • Fix wrong args (6db14ab)
  • Update pixelize.py (e2b09df)
  • Update README for ComfyUI implementation (175e052)
  • Update node implementation (1b0a131)
  • 0.1.3 (e03c8af)
  • Formatting (60f7e84)
  • ComfyUI Node initial implementation (bb3799a)
  • return weight when applicable (3c2a5f2)
  • Resolve antialias result (883fce4)
  • Support return oe/weight in pixelize (aa75441)
  • Create init.py (060aeca)
  • Update the intro in demo-gr (b0f4ef9)
  • Add dtype selection (948a367)
  • remove unused import (b56a098)
  • Update demo-torch.py (adcd0ab)
  • Update Readme (5299c22)
  • Update env.py (abac649)
  • Update k_centroid.py (3595ddd)
  • Update demo-gr.py (0b7378f)
  • Update README.md (94e09f3)
  • better kmeans with better initial centroids (97be187)
  • Use compile for repeat-kmeans (890423a)
  • update demo for weighted kmeans and repeat kmeans (de3aff9)
  • implement weighted and repeat kmeans (b837029)
  • avoid .view (954f664)
  • Update gradio client (80e6d43)
  • Add gradio clients (0d3bb1c)
  • Update outline.py (f122485)
  • Add more downscale mode (5b3295e)
  • seperate resize into another function (e3400dd)
  • Formatting (11d8380)
  • modify some details (80c18af)
  • Create logger.py (025343d)
  • better naming (beee109)
  • move to legacy (421f3b0)
  • Update utils.py (28a493f)
  • remove debug print (d9fe29f)
  • switch compile/not compile at runtime (cd870e7)
  • Move old implementation to "legacy" (dd316f3)
  • refine some design choice (a99c9a6)
  • use lossless webp for better result (81161ec)
  • Fix some small details (dba5eed)
  • update demo images (5381036)
  • Avoid non-deterministic kmeans (7c5314d)
  • Fix the k_centroid implementation (9a90412)
  • new demo for batched, k-centroid (600dfb6)
  • Finish the batched-torch-k-centroid downsample (3a73956)
  • move batched kmeans to utils (9bdabcb)
  • Fully batched pipeline done (a135c73)
  • Update utils.py (cb3d11b)
  • Fix odd patch size for outline expansion (5deb943)
  • Fix the batched contrast based downsample (f689ec9)
  • remove all pngs (fd5e26a)
  • use webp instead of png (574f1e3)
  • Formalize the demo/benchmark (3da072e)
  • Formatting (2c8109d)
  • Update demos for FP16 and batch (aa7ff72)
  • Support fp16 (cd91e05)
  • Support batch processing (d6e5e9f)
  • update demo (6afb1a1)
  • update demo for dithering (884df72)
  • implementation of dithering (5b0f5a3)
  • 0.1.1 (4067e81)
  • Update demos (9683d0d)
  • remove one file impl (f80dadc)
  • Move torch impl of pixelization into single file (105b4ee)
  • Add kmeans in torch impl (3cd0ada)
  • Create utils.py (eb798c2)
  • Move torch impl of outline expansion algo into single file (e9c5ed6)
  • Create init.py (4a0fa00)
  • Move the contrast downsample into single file (b4df16a)
  • Implement wavelet color match in torch (ed66f3c)
  • Update README.md (080f419)
  • Move source code into src/ (e282fe3)
  • Update demos (26d9fbc)
  • Refine readme (1abcfec)
  • Update README.md (92a9be3)
  • Update the outline expansion informations (36114df)
  • Update init.py (a0c9de7)
  • refine the implementation of pure torch one (b4079ca)
  • Update .gitignore (370a286)
  • Full torch implementation (6993834)
  • Optimizations (9e80eec)
  • Add demo for Weighted-Kmeans (b5b10ac)
  • Formatting (af2f4c0)
  • Weighted KMeans (caa17c3)
  • custom color match level and more quant method (a1646da)
  • Fix setup problem, 0.0.10 (8728764)
  • Bug fixes (48903f6)
  • 0.0.9 (0269202)
  • Fix padding problem (1e3a32a)
  • Fix install import problem (141f2a1)
  • Better isiterable(maybe) (b8a05a0)
  • better naming to avoid conflict with built-in (f210244)
  • use _ for unused var (8149315)
  • 0.0.7 (ff37470)
  • formatting (7ac3ef0)
  • Use correct ratio (e8b07e8)
  • 0.0.6 (36c3cf8)
  • support target_size=(W, H) in cli (07771ba)
  • Fix import order (71400e0)
  • Add isiterable utils (9c61148)
  • Add (W, H) target size support (e5cb476)
  • Add color quant (93063d3)
  • Update README.md (9b33f87)
  • Update README.md (d0add27)
  • Add license info in setup.py (4a15093)
  • 0.0.5 (1f8dfff)
  • add more command(outline) (1bb34dc)
  • Add edge demo (cf5e3c1)
  • Update demo.py (8cadc38)
  • Update README.md (c9cdbc6)
  • Update README.md (0e62898)
  • Update README.md (5836f7e)
  • Update pixelize.py (73690d0)
  • fix the python example (768731f)
  • fix default mode (82e0c47)
  • Fix bicubic downscale with PIL (83f36dc)
  • Update contrast_based.py (2a69a2c)
  • Update README.md (ea71d6c)
  • Create demo.py (246f2f9)
  • Add example of improving other method (608e77d)
  • Update README.md (dbb7722)
  • Add more example (af61a74)
  • Fix color-match arg and add no_downscale (cf63b23)
  • 0.0.4 (88cad97)
  • Use short name for contrast-based downscaling (5068101)
  • implement contrast-based downscale (7161cfe)
  • Better implementation of apply_chunk (a8df9eb)
  • update demo imgs (08f2860)
  • Update README.md (1f33476)
  • Update demo (9594c2e)
  • Add conventional algo (48a5503)
  • Update README.md (de1fb99)
  • Update README.md (856676d)
  • algorithm name correction (cb4bacc)
  • Update README.md (61cea27)
  • Update README.md (5aecb30)
  • Update example images (9d5e1c4)
  • Cli implementation for choosing downscale method (1c68630)
  • Add license for k-centroid algo (e36fcb8)
  • implement k-centroid downscale method (e9d0fb5)
  • 0.0.3 (8dcaf0e)
  • formatting (4992071)
  • Update cli.py (a70c662)
  • Update test2.png (c108b72)
  • Example with new implementation (6d233ad)
  • Use torch.Fold/Unfold for faster speed (79eb0c1)
  • Update README.md (80b9f82)
  • update readme and example (9aaee76)
  • formatting (90a2ad3)
  • Update README.md (b5d125d)
  • cli support (a352c88)
  • Update README.md (1bb72c0)
  • first commit (f583f41)
Assets 4
Loading