helloeditingmaster commited on
Commit
e204c4d
·
verified ·
1 Parent(s): 4619cfa

v0.1.0 Windows: BiRefNet copies without DeformConv (birefnet_lite_win, birefnet_win)

Browse files
README.md CHANGED
@@ -20,7 +20,7 @@ The app runs fully offline after this download. These files are not meant to be
20
 
21
  | Folder | Model | License | Original source |
22
  |---|---|---|---|
23
- | `birefnet_lite`, `birefnet`, `birefnet_lite_2k` | BiRefNet lite / full / lite-2K | MIT ([LICENSE-BiRefNet-MIT.txt](LICENSE-BiRefNet-MIT.txt)) | Peng Zheng et al., "Bilateral Reference for High-Resolution Dichotomous Image Segmentation" — https://github.com/ZhengPeng7/BiRefNet (weights: huggingface.co/ZhengPeng7/BiRefNet, BiRefNet_lite, BiRefNet_lite-2K) |
24
  | `sam2.1_tiny`, `sam2.1_small`, `sam2.1_base_plus`, `sam2.1_tiny_512`, `sam2.1_small_768` | SAM 2.1 hiera-tiny / small / base-plus | Apache-2.0 ([LICENSE-SAM2-Apache-2.0.txt](LICENSE-SAM2-Apache-2.0.txt)) | Meta AI, "SAM 2: Segment Anything in Images and Videos" — https://github.com/facebookresearch/sam2 |
25
 
26
  ## Changes from the originals
@@ -28,6 +28,9 @@ The app runs fully offline after this download. These files are not meant to be
28
  - Converted from the official PyTorch weights to ONNX by Master Mask's export tool.
29
  - BiRefNet: two tensor reshapes are written differently (same maths) so the model runs on Apple's CoreML;
30
  `birefnet_lite_2k` is exported at two fixed input sizes (landscape and portrait).
 
 
 
31
  - SAM 2.1: split into image encoder, mask decoder, memory encoder and memory attention (for video tracking);
32
  `*_512` / `*_768` are the same weights exported at a lower working resolution. Small constant tensors are stored
33
  in `constants.npz` / `video_constants.npz`.
 
20
 
21
  | Folder | Model | License | Original source |
22
  |---|---|---|---|
23
+ | `birefnet_lite`, `birefnet`, `birefnet_lite_2k`, `birefnet_lite_win`, `birefnet_win` | BiRefNet lite / full / lite-2K | MIT ([LICENSE-BiRefNet-MIT.txt](LICENSE-BiRefNet-MIT.txt)) | Peng Zheng et al., "Bilateral Reference for High-Resolution Dichotomous Image Segmentation" — https://github.com/ZhengPeng7/BiRefNet (weights: huggingface.co/ZhengPeng7/BiRefNet, BiRefNet_lite, BiRefNet_lite-2K) |
24
  | `sam2.1_tiny`, `sam2.1_small`, `sam2.1_base_plus`, `sam2.1_tiny_512`, `sam2.1_small_768` | SAM 2.1 hiera-tiny / small / base-plus | Apache-2.0 ([LICENSE-SAM2-Apache-2.0.txt](LICENSE-SAM2-Apache-2.0.txt)) | Meta AI, "SAM 2: Segment Anything in Images and Videos" — https://github.com/facebookresearch/sam2 |
25
 
26
  ## Changes from the originals
 
28
  - Converted from the official PyTorch weights to ONNX by Master Mask's export tool.
29
  - BiRefNet: two tensor reshapes are written differently (same maths) so the model runs on Apple's CoreML;
30
  `birefnet_lite_2k` is exported at two fixed input sizes (landscape and portrait).
31
+ - BiRefNet for Windows (`birefnet_lite_win`, `birefnet_win`): the same models with each DeformConv
32
+ (deformable convolution) written as standard operations (GridSample, MatMul), because the DirectML build of
33
+ ONNX Runtime has no DeformConv. Same maths: largest output difference 1.7e-5.
34
  - SAM 2.1: split into image encoder, mask decoder, memory encoder and memory attention (for video tracking);
35
  `*_512` / `*_768` are the same weights exported at a lower working resolution. Small constant tensors are stored
36
  in `constants.npz` / `video_constants.npz`.
birefnet_lite_win/info.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "kind": "birefnet",
3
+ "inputs": [
4
+ {
5
+ "file": "model.onnx",
6
+ "hw": [
7
+ 1024,
8
+ 1024
9
+ ]
10
+ }
11
+ ],
12
+ "windows_copy_of": "birefnet_lite"
13
+ }
birefnet_lite_win/model.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:361da7a3edd014bea96402be9f965b4342e420d882442c80644906157f543336
3
+ size 181477493
birefnet_win/info.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "kind": "birefnet",
3
+ "inputs": [
4
+ {
5
+ "file": "model.onnx",
6
+ "hw": [
7
+ 1024,
8
+ 1024
9
+ ]
10
+ }
11
+ ],
12
+ "skip_providers": {
13
+ "model.onnx": [
14
+ "CoreMLExecutionProvider",
15
+ "DmlExecutionProvider"
16
+ ]
17
+ },
18
+ "windows_copy_of": "birefnet"
19
+ }
birefnet_win/model.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ae7220144d2a829007ae210acc34d9d593bc8d4d97d03ab3bc45c084e99b7d1
3
+ size 930456813
manifest.json CHANGED
@@ -29,6 +29,22 @@
29
  "size": 181080713,
30
  "sha256": "12ce2e27f243838697a12dc7174595b23322de0b97bfa0d00c6d9162af8dc435"
31
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
32
  "sam2.1_base_plus/constants.npz": {
33
  "size": 2394,
34
  "sha256": "d3885e44941cbd8b9313e46d04b904b0bc83c61ef5c54d96885a98332b4627a2"
 
29
  "size": 181080713,
30
  "sha256": "12ce2e27f243838697a12dc7174595b23322de0b97bfa0d00c6d9162af8dc435"
31
  },
32
+ "birefnet_lite_win/info.json": {
33
+ "size": 170,
34
+ "sha256": "40435b2de7e2a60ce5e5437011e29adb59527cacaca7cc0828786bca70b77e38"
35
+ },
36
+ "birefnet_lite_win/model.onnx": {
37
+ "size": 181477493,
38
+ "sha256": "361da7a3edd014bea96402be9f965b4342e420d882442c80644906157f543336"
39
+ },
40
+ "birefnet_win/info.json": {
41
+ "size": 280,
42
+ "sha256": "6430f0fbcb7a672366bd43669e56bf2d8de604a64386249dede36a6a2615b9d7"
43
+ },
44
+ "birefnet_win/model.onnx": {
45
+ "size": 930456813,
46
+ "sha256": "0ae7220144d2a829007ae210acc34d9d593bc8d4d97d03ab3bc45c084e99b7d1"
47
+ },
48
  "sam2.1_base_plus/constants.npz": {
49
  "size": 2394,
50
  "sha256": "d3885e44941cbd8b9313e46d04b904b0bc83c61ef5c54d96885a98332b4627a2"