Flor commited on
Commit
c4e6f55
·
verified ·
1 Parent(s): 62d81da

card: update from fastforks 1cbf1671

Browse files
Files changed (1) hide show
  1. README.md +100 -0
README.md ADDED
@@ -0,0 +1,100 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: "fastforks"
3
+ pipeline_tag: "image-to-image"
4
+ license: "other"
5
+ license_name: "see-model-card"
6
+ license_link: "https://github.com/huggingface/controlnet_aux/blob/b1aac318b15c4e65a21bc516dc236ad4bf593d46/src/controlnet_aux/pidi/LICENSE"
7
+ tags:
8
+ - "edges"
9
+ - "edge-detection"
10
+ - "line-art"
11
+ - "controlnet-preprocessor"
12
+ - "cuda"
13
+ - "pytorch"
14
+ - "fastforks"
15
+ - "fastforks-card"
16
+ ---
17
+ # PiDiNet soft edges (fastforks)
18
+
19
+ Soft edge preprocessor of ControlNet 1.1 (controlnet_aux `PidiNetDetector`, ComfyUI's "PiDiNet Soft-Edge Lines"): pixel difference networks (Su et al., ICCV 2021), a 0.7 M-parameter edge detector of difference convolutions, here converted to plain depthwise convolutions (the authors' re-parameterization). Research use only.
20
+
21
+ This repo is the model card only: it holds no weights. fastforks downloads the authors' weights from the links below and checks them against the SHA-256 in the model card. Code: <https://github.com/kleinicke/fastforks>. Kernels for the `kernels` backend: <https://huggingface.co/fastforks/kernels>. Original code: <https://github.com/huggingface/controlnet_aux> at `b1aac318b15c`.
22
+
23
+ Author weight repos: <https://huggingface.co/lllyasviel/Annotators>.
24
+
25
+ Task: edges. Tier: experimental. Outputs: `edges`. Backends: `portable`.
26
+
27
+ Input: one RGB image. Output: a line / edge map `[H, W]` in [0, 1] at the input size (1 = line), the ControlNet preprocessors (line art, HED, PiDiNet, TEED / Anyline); M-LSD also returns straight line segments.
28
+
29
+ ## Variants
30
+
31
+ Weights are the authors' files (not hosted here). SHA-256 shortened to 12 hex digits; the full hashes are in [`card.toml`](https://github.com/kleinicke/fastforks/blob/main/src/fastforks/edges/models/pidinet/card.toml).
32
+
33
+ | variant | weights (author link) | SHA-256 | weights licence | commercial use |
34
+ |---|---|---|---|---|
35
+ | `pidinet@1` | [table5_pidinet.pth](https://huggingface.co/lllyasviel/Annotators/resolve/982e7edaec38759d914a963c48c4726685de7d96/table5_pidinet.pth) | `80860ac26725` | LicenseRef-research-only (assumed from code) | **no** |
36
+
37
+ ## Licence
38
+
39
+ Original-code strings, recorded as they are (research-use package; check them before any other use). Code: LicenseRef-research-only. Licence files: `LICENSE`.
40
+
41
+ Code licence: LicenseRef-research-only; the authors do not state a separate weights licence.
42
+
43
+ Training data: Not stated for these checkpoints (lllyasviel/Annotators, ControlNet 1.1). The file name matches the paper's table 5 BSDS500 model..
44
+
45
+ Licence evidence: https://github.com/huggingface/controlnet_aux/blob/b1aac318b15c4e65a21bc516dc236ad4bf593d46/src/controlnet_aux/pidi/LICENSE ; https://github.com/hellozhuo/pidinet/blob/master/LICENSE ; https://huggingface.co/lllyasviel/Annotators
46
+
47
+ ## Input size
48
+
49
+ any H x W >= 32, padded to a multiple of 32
50
+ Tested up to (portable, fp16, RTX 5060 Ti 16 GB): `pidinet` 1080x1920 (0.3 GiB peak allocated, 0.4 reserved)
51
+
52
+ ## Speed and memory
53
+
54
+ ### Edges, synthetic, 1080x1920
55
+
56
+ | model | time (ms): original | portable | peak memory (MiB, allocated / reserved): original | portable |
57
+ | --- | --- | --- | --- | --- |
58
+ | `pidinet` | 35.9 | 13.1 | 339 / 368 | 328 / 372 |
59
+
60
+ ### Startup and first call
61
+
62
+ | model | load (s): original | portable | kernels | first call (s): original | portable | kernels |
63
+ | --- | --- | --- | --- | --- | --- | --- |
64
+ | `pidinet` | 0.28 | 0.16 | – | 0.06 | 0.01 | – |
65
+
66
+ ### torch.compile and TensorRT
67
+
68
+ | model | input | time: portable | compile | TensorRT fp16 | kernels | preparation: compile (cold first call) | ONNX + engine build | peak MiB (alloc / reserved): portable | compile | kernels | TensorRT (engine + torch) |
69
+ | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
70
+ | `pidinet` | 1024x1024 | 6.59 ms | 5.66 ms | 4.50 ms | – | 12.4 s | 20.4 s | 186 / 228 | 166 / 190 | – | 187 / 200 |
71
+
72
+ Fixed-size cells: time (ms) / peak allocated / peak reserved memory (MiB).
73
+
74
+ Hardware, protocol and caveats: [benchmarks](https://github.com/kleinicke/fastforks/blob/main/docs/benchmarks.md) (RTX 5060 Ti 16 GB, sm_120). Original = the original code named above at the precision of portable (fp16, or `(fp32)` where fp16 is unsupported or not finite); original fp32 = its default fp32 run. Peak memory is the torch allocator's.
75
+
76
+ ## Accuracy
77
+
78
+ ### Parity with the original code
79
+
80
+ Every variant is checked against reference outputs of the original code at its pinned commit. CPU, float32: max |diff| <= 0.002 vs the original. GPU, float16 (the default), portable and kernels: mean |diff| <= 0.02 and p99 <= 0.25 (absolute error of the output `edges`).
81
+
82
+ ## Usage
83
+
84
+ ```python
85
+ from fastforks import edges
86
+
87
+ # non-commercial weights: run with FASTFORKS_NONCOMMERCIAL=1 in the environment
88
+ model = edges.load("pidinet") # mode="auto": the preferred backend from recorded measurements
89
+ result = model("photo.jpg")
90
+ result.edges # [H, W] float32 in [0, 1], 1 = line, input resolution
91
+ result.backend # what ran, e.g. Backend(mode='portable', device='cuda:0', dtype='float16')
92
+ ```
93
+
94
+ Install with `pip install fastforks` (see <https://github.com/kleinicke/fastforks>). Other variants: `pidinet`.
95
+
96
+ ## Kernels backend
97
+
98
+ No kernels backend: the converted network is plain convolutions (cuDNN); no custom kernel measured to pay off.
99
+
100
+ <!-- staged from fastforks 1cbf1671 (2026-11-15) -->