Hugging Face's logo

WeReCooking
/
sapiens2-onnx

Depth Estimation
sapiens2
ONNX
cpu
segmentation
pose-estimation
normal-estimation
Model card Files
xet
 Community
Nekochu commited on
Commit
2d6b460
·
verified ·
1 Parent(s): 7c9cce7

Initial commit

Browse files
Files changed (32) hide show
  1. .gitattributes +15 -0
  2. README.md +57 -0
  3. normal/normal_0.4b_fp32.onnx +3 -0
  4. normal/normal_0.4b_fp32.onnx.data +3 -0
  5. normal/normal_0.8b_fp32.onnx +3 -0
  6. normal/normal_0.8b_fp32.onnx.data +3 -0
  7. normal/normal_1b_fp32.onnx +3 -0
  8. normal/normal_1b_fp32.onnx.data +3 -0
  9. normal/normal_5b_int8.onnx +3 -0
  10. normal/normal_5b_int8.onnx.data +3 -0
  11. pointmap/pointmap_0.4b_fp32.onnx +3 -0
  12. pointmap/pointmap_0.4b_fp32.onnx.data +3 -0
  13. pointmap/pointmap_0.8b_fp32.onnx +3 -0
  14. pointmap/pointmap_0.8b_fp32.onnx.data +3 -0
  15. pointmap/pointmap_1b_fp32.onnx +3 -0
  16. pointmap/pointmap_1b_fp32.onnx.data +3 -0
  17. pointmap/pointmap_5b_int8.onnx +3 -0
  18. pointmap/pointmap_5b_int8.onnx.data +3 -0
  19. pose/pose_0.4b_fp32.onnx +3 -0
  20. pose/pose_0.4b_fp32.onnx.data +3 -0
  21. pose/pose_0.8b_fp32.onnx +3 -0
  22. pose/pose_0.8b_fp32.onnx.data +3 -0
  23. pose/pose_1b_fp32.onnx +3 -0
  24. pose/pose_1b_fp32.onnx.data +3 -0
  25. seg/seg_0.4b_fp16.onnx +3 -0
  26. seg/seg_0.4b_fp16.onnx.data +3 -0
  27. seg/seg_0.8b_fp32.onnx +3 -0
  28. seg/seg_0.8b_fp32.onnx.data +3 -0
  29. seg/seg_1b_fp32.onnx +3 -0
  30. seg/seg_1b_fp32.onnx.data +3 -0
  31. seg/seg_5b_int8.onnx +3 -0
  32. seg/seg_5b_int8.onnx.data +3 -0
.gitattributes CHANGED
@@ -33,3 +33,18 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ seg/seg_0.4b_fp16.onnx.data filter=lfs diff=lfs merge=lfs -text
37
+ seg/seg_0.8b_fp32.onnx.data filter=lfs diff=lfs merge=lfs -text
38
+ seg/seg_1b_fp32.onnx.data filter=lfs diff=lfs merge=lfs -text
39
+ seg/seg_5b_int8.onnx.data filter=lfs diff=lfs merge=lfs -text
40
+ normal/normal_0.4b_fp32.onnx.data filter=lfs diff=lfs merge=lfs -text
41
+ normal/normal_0.8b_fp32.onnx.data filter=lfs diff=lfs merge=lfs -text
42
+ normal/normal_1b_fp32.onnx.data filter=lfs diff=lfs merge=lfs -text
43
+ normal/normal_5b_int8.onnx.data filter=lfs diff=lfs merge=lfs -text
44
+ pointmap/pointmap_0.4b_fp32.onnx.data filter=lfs diff=lfs merge=lfs -text
45
+ pointmap/pointmap_0.8b_fp32.onnx.data filter=lfs diff=lfs merge=lfs -text
46
+ pointmap/pointmap_1b_fp32.onnx.data filter=lfs diff=lfs merge=lfs -text
47
+ pointmap/pointmap_5b_int8.onnx.data filter=lfs diff=lfs merge=lfs -text
48
+ pose/pose_0.4b_fp32.onnx.data filter=lfs diff=lfs merge=lfs -text
49
+ pose/pose_0.8b_fp32.onnx.data filter=lfs diff=lfs merge=lfs -text
50
+ pose/pose_1b_fp32.onnx.data filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: other
3
+ tags:
4
+ - onnx
5
+ - sapiens2
6
+ - cpu
7
+ - segmentation
8
+ - pose-estimation
9
+ - normal-estimation
10
+ - depth-estimation
11
+ ---
12
+
13
+ # Sapiens2 ONNX
14
+
15
+ CPU-friendly ONNX exports of Meta's `facebook/sapiens2-*`. 15 task heads across 4 tasks and 4 sizes.
16
+
17
+ ## Folder layout
18
+
19
+ Each task has its own folder. Each model is split into a small `.onnx` graph file plus a `.onnx.data` external sidecar (both must live in the same directory at download time).
20
+
21
+ | Task | 0.4b | 0.8b | 1b | 5b |
22
+ |---|---|---|---|---|
23
+ | seg | `seg/seg_0.4b_fp16.onnx` (777 MB, fp16) | `seg/seg_0.8b_fp32.onnx` (3.3 GB) | `seg/seg_1b_fp32.onnx` (5.9 GB) | `seg/seg_5b_int8.onnx` (5.2 GB) |
24
+ | normal | `normal/normal_0.4b_fp32.onnx` (1.7 GB) | `normal/normal_0.8b_fp32.onnx` (3.5 GB) | `normal/normal_1b_fp32.onnx` (6.2 GB) | `normal/normal_5b_int8.onnx` (6.1 GB) |
25
+ | pointmap | `pointmap/pointmap_0.4b_fp32.onnx` (2.0 GB) | `pointmap/pointmap_0.8b_fp32.onnx` (3.9 GB) | `pointmap/pointmap_1b_fp32.onnx` (6.5 GB) | `pointmap/pointmap_5b_int8.onnx` (6.2 GB) |
26
+ | pose | `pose/pose_0.4b_fp32.onnx` (1.6 GB) | `pose/pose_0.8b_fp32.onnx` (3.4 GB) | `pose/pose_1b_fp32.onnx` (6.1 GB) | not shipped |
27
+
28
+ Cosine similarity vs the PyTorch reference is 0.999 or better on every shipped file.
29
+
30
+ ## Precision notes
31
+
32
+ * seg-0.4b is fp16 (50 percent smaller than fp32 and verified cos 0.99999)
33
+ * 0.4b/0.8b/1b for normal, pointmap, pose are fp32. Naive fp16 cast produces NaN (normal L2-normalize divides near zero) or cos around 0.7 (pointmap metric scale, pose sigmoid heatmaps saturate)
34
+ * 5B variants are INT8 (per-channel symmetric, MatMulIntegerToFloat)
35
+ * pose-5b is not shipped (the int8 quantize attempt did not complete on the available hardware)
36
+
37
+ ## Inference
38
+
39
+ ```python
40
+ import numpy as np
41
+ import onnxruntime as ort
42
+ from huggingface_hub import hf_hub_download
43
+
44
+ # Download both the .onnx graph and the .onnx.data sidecar side by side
45
+ for fn in ("seg/seg_0.4b_fp16.onnx", "seg/seg_0.4b_fp16.onnx.data"):
46
+ hf_hub_download(repo_id="WeReCooking/sapiens2-onnx", filename=fn, local_dir=".")
47
+
48
+ sess = ort.InferenceSession("seg/seg_0.4b_fp16.onnx", providers=["CPUExecutionProvider"])
49
+ # Input expects (N, 3, 1024, 768) fp32 BGR mean-subtracted preprocessed tensor
50
+ out = sess.run(None, {"input": preprocessed})
51
+ ```
52
+
53
+ For a standalone CLI without sapiens2 or PyTorch, see `app.py onnx ...` in the source Space `WeReCooking/sapiens2-cpu`.
54
+
55
+ ## License
56
+
57
+ Same as upstream `facebook/sapiens2-*`.
normal/normal_0.4b_fp32.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a6a98218a68ad85c9f3bee5ab193e4cbc34d82c3ee963399479c8fd52a293fa0
3
+ size 3764421
normal/normal_0.4b_fp32.onnx.data ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e786a5caba89d7df98bafa02980c1b2622e2e4f94ccaa6aae2c955ed7b8ce08b
3
+ size 1813839872
normal/normal_0.8b_fp32.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:04fbfdd8d825ced3bed76e65b4fd060e7badaf1793f409bbceed6b307b8a689b
3
+ size 5062864
normal/normal_0.8b_fp32.onnx.data ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a8864316fffbaefcf406bf0c79c5b5677de2fa1a8c309b1e64cf65b4d35a97b3
3
+ size 3538878464
normal/normal_1b_fp32.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b05e410dcbc6ac005043e543b80708c903cd5eecd23573f9a12f39b84487126
3
+ size 6384132
normal/normal_1b_fp32.onnx.data ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:71f857f5b5fcb4ec417eff0709e7b282412402412cc8726b391dac2f4669464e
3
+ size 6157762560
normal/normal_5b_int8.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:225dd119e0619ebc3b218dca6d7a54320e5ef02caefc8b84bfe408d068c6c599
3
+ size 16435117
normal/normal_5b_int8.onnx.data ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d7185b19b7b6121cd5a5f81bd21826a442bb7eae3ee126404a434edd761ed8a
3
+ size 6077154301
pointmap/pointmap_0.4b_fp32.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:edd61f29cec8b9b6357056d0c93bad09376ead463466c3ef99c18bb263ccd6c8
3
+ size 3829289
pointmap/pointmap_0.4b_fp32.onnx.data ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b09043fb2d65a6ff00849badd543627660476bf68e771e082f02c9dc13d6d4f2
3
+ size 2115174400
pointmap/pointmap_0.8b_fp32.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9141804c5d560847a82cc52dad1a5793a68659869b2f01f78cd7dafa6152e7ca
3
+ size 5135100
pointmap/pointmap_0.8b_fp32.onnx.data ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4e1fe86174568e7b1deecc75a247ec0239a0a59d2fb0585be74a0b800981f6fd
3
+ size 3870031872
pointmap/pointmap_1b_fp32.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a2c1323d7d6bfee2917555514f18abceee3e3405963f421530e9472a518481a3
3
+ size 6463539
pointmap/pointmap_1b_fp32.onnx.data ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:df9c4313af95dcf3c0f6b754549cedc30d4aa917cad45adbdd7cc3716085fb0e
3
+ size 6518800384
pointmap/pointmap_5b_int8.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f8991b83b92e8e84a5f3b88722b3d3bfe2419f300a95ea1a5c2a5b9a5398ac7f
3
+ size 16521512
pointmap/pointmap_5b_int8.onnx.data ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ec675354c3a018cc491872e6b3fffaf0971796ee88666ee962a2b465a70d8ae
3
+ size 6195479548
pose/pose_0.4b_fp32.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a995a2c503d41d3323b9e6f2218197bad9a2b51533f90e511ca551c185fc361b
3
+ size 3771752
pose/pose_0.4b_fp32.onnx.data ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d1d45dc8db765a2a8dc04a1e6b5f288be103da9ec99ad404d55c9d0ae57d33be
3
+ size 1703149568
pose/pose_0.8b_fp32.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4ab377312ef6973dd92e1330ba48dadb49950b46520450d7173ff7f490e28e6a
3
+ size 5087421
pose/pose_0.8b_fp32.onnx.data ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6c829320ce456925e371afc8938a794dfc8f7ae95f67e99a1e0d5ac2e4062bae
3
+ size 3395354624
pose/pose_1b_fp32.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:273c7606612ee269a38fc4b33f0b04c0e665883edcf1762fa676f0ebd681b113
3
+ size 6426602
pose/pose_1b_fp32.onnx.data ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:caf5e3fd7b01a74a3ad4005c9e0aad0bb4918727a9cb7dceb6ab9bb76f85bc99
3
+ size 6079578112
seg/seg_0.4b_fp16.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b2cf9f8076e42c6994383a974b37ce866bfbd055432004981d292bc2e6b4ca19
3
+ size 3765586
seg/seg_0.4b_fp16.onnx.data ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9845fe46d00ec728765d45775c42a44d55d8f8943c07772f470043498b0a7560
3
+ size 813395582
seg/seg_0.8b_fp32.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6674eed65c8350cddd0ede3ff1f0a604c935586bd2bbba5ea8f81bfa3f9b61be
3
+ size 4986578
seg/seg_0.8b_fp32.onnx.data ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab2154cd824b1c905f030467577c8e11ece1f8385c27512fefa81c9beeb81716
3
+ size 3310682112
seg/seg_1b_fp32.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:25c3dda389316b87ac9e1a80254af337155bc3e0d7e9af74393d9c392bc77db2
3
+ size 6297140
seg/seg_1b_fp32.onnx.data ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:63e267ab8bd7631abe3c227e6d4bb2d8902902f591124836777b7d16434f62db
3
+ size 5883691008
seg/seg_5b_int8.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:366c320c66574795baa4e56c77463aa6753da7c92180a4789254a9b4e9634db9
3
+ size 16526759
seg/seg_5b_int8.onnx.data ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:623f89266c2d563f31e9f47d75ff804b905686c1c947d26a8e77e34bebf059ce
3
+ size 5170038526