Rehost regression_emg2pose.ckpt with braindecode parameter names

README.md

@@ -0,0 +1,41 @@
+---
+library_name: braindecode
+license: cc-by-nc-sa-4.0
+tags: [emg, semg, hand-pose, pose-estimation, braindecode, emg2pose]
+pipeline_tag: other
+---
+
+# emg2pose — emg2pose Regression baseline
+
+Meta's released `regression_emg2pose.ckpt` from the
+[emg2pose benchmark](https://github.com/facebookresearch/emg2pose), rehosted with
+[braindecode](https://braindecode.org)'s parameter names:
+
+```python
+from braindecode.models import VEMG2Pose
+
+model = VEMG2Pose.from_pretrained("braindecode/EMG2Pose-emg2pose")
+```
+
+16-channel sEMG at 2 kHz in, 20 joint angles per sample out. The encoder uses **valid**
+convolutions and consumes a left context of **1790** samples, so windows must be longer
+than that; the paper trains on 11,790 (10,000 + 1790).
+
+## Configuration
+
+`decoder="mlp"`, `parameterization="position"`. The stateless MLP decoder reads the pose out directly at each step, still conditioned on the previous pose. Trained without a ground-truth initial pose.
+
+## Provenance
+
+Not retrained — the authors' checkpoint with parameter names rewritten. Loaded into
+`braindecode.models.VEMG2Pose` and run against the reference `emg2pose.pose_modules`
+implementation on the same input, the outputs are **bit-identical** (max absolute
+difference 0.0), with all 68 tensors mapped.
+
+## License
+
+Weights are Meta's, under **CC BY-NC-SA 4.0** (non-commercial, share-alike), carried
+over unchanged. The UmeTrack hand model behind the labels is CC BY-NC 4.0.
+
+- Salter et al. (2024), *emg2pose*, NeurIPS Datasets and Benchmarks.
+  [arXiv:2412.02725](https://arxiv.org/abs/2412.02725)

config.json

@@ -0,0 +1,46 @@
+{
+  "n_outputs": 20,
+  "n_chans": 16,
+  "chs_info": null,
+  "n_times": 11790,
+  "input_window_seconds": null,
+  "sfreq": 2000.0,
+  "encoder_channels": 256,
+  "feature_dim": 64,
+  "hidden_size": 512,
+  "lstm_layers": 2,
+  "decoder": "mlp",
+  "decoder_hidden_sizes": [
+    512,
+    512
+  ],
+  "stem_kernel_sizes": [
+    11,
+    5
+  ],
+  "stem_strides": [
+    5,
+    2
+  ],
+  "tds_subsample_kernels": [
+    17,
+    9
+  ],
+  "tds_subsample_strides": [
+    4,
+    2
+  ],
+  "tds_kernel_widths": [
+    9,
+    5
+  ],
+  "tds_blocks": 2,
+  "tds_channels": 16,
+  "rollout_rate": 50.0,
+  "num_position_steps": 500,
+  "parameterization": "position",
+  "output_scalar": 0.01,
+  "activation": "torch.nn.modules.activation.LeakyReLU",
+  "drop_prob": 0.0,
+  "braindecode_version": "1.8.0dev0"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e18231ffe69e3f66bd6672313a7762c9a81fb4ddaf033de2380361d1eb84cf9
+size 11817144

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ea6e462bfd57963e183a74cb7ac20975c18e5b1c2154be02828dc119c2277d7
+size 11836858