cozyclay 1.9.0 → 2.0.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (286) hide show
  1. package/CHANGELOG.md +171 -0
  2. package/bin/agent/agent-routes.mjs +74 -70
  3. package/bin/agent/agent-runner.mjs +10 -3
  4. package/bin/agent/motion-runtime.mjs +32 -260
  5. package/bin/agent/providers.mjs +46 -3
  6. package/bin/agent/session-store.mjs +41 -5
  7. package/bin/agent/studio-prompt.mjs +1 -1
  8. package/bin/agent/studio-tools.mjs +48 -11
  9. package/bin/cozyclay.mjs +26 -8
  10. package/bin/live/cli.mjs +22 -3
  11. package/bin/mcp-runtime.mjs +35 -20
  12. package/bin/telemetry-state.mjs +5 -2
  13. package/bin/update-check.mjs +15 -0
  14. package/dist/app/index.html +6 -6
  15. package/dist/assets/analytics-MVZOEW8O.js +1 -0
  16. package/dist/assets/app-C1sdVbg4.css +1 -0
  17. package/dist/assets/app-CA55tcI3.js +4876 -0
  18. package/dist/assets/{demo-B-eg1CT8.js → demo-ByeYp-dY.js} +1 -1
  19. package/dist/assets/{first-shot-handoff-CGgnvNu8.js → first-shot-handoff-DgT28ZiV.js} +1 -1
  20. package/dist/assets/{landing-IbRp3FnQ.js → landing-BaXkpFip.js} +1 -1
  21. package/dist/assets/shot-prompt-CbqWcWYi.css +1 -0
  22. package/dist/assets/shot-prompt-Yrt99wZx.js +17 -0
  23. package/dist/assets/{ticket-BgIzOBOf.js → ticket-BcmsnHn5.js} +1 -1
  24. package/dist/assets/workflow-DRdwoDBT.css +1 -0
  25. package/dist/assets/workflow-dNPGY9xz.js +179 -0
  26. package/dist/cozyclay-package.json +1 -1
  27. package/dist/fonts/IBM-Plex-Sans-OFL.txt +92 -0
  28. package/dist/fonts/JetBrains-Mono-OFL.txt +93 -0
  29. package/dist/fonts/README.md +13 -9
  30. package/dist/fonts/ibm-plex-sans-400-latin.woff2 +0 -0
  31. package/dist/fonts/ibm-plex-sans-500-latin.woff2 +0 -0
  32. package/dist/fonts/ibm-plex-sans-600-latin.woff2 +0 -0
  33. package/dist/fonts/jetbrains-mono-400-latin.woff2 +0 -0
  34. package/dist/fonts/jetbrains-mono-500-latin.woff2 +0 -0
  35. package/dist/index.html +7 -7
  36. package/dist/privacy/index.html +13 -8
  37. package/dist/sitemap.xml +6 -6
  38. package/dist/workflow/index.html +5 -5
  39. package/mcp/LIVE-PROTOCOL.md +1 -1
  40. package/mcp/live-hub.mjs +35 -93
  41. package/mcp/mesh-file.mjs +48 -0
  42. package/mcp/server.mjs +6 -77
  43. package/mcp/tool-handlers.mjs +224 -377
  44. package/package.json +2 -1
  45. package/src/App.jsx +1273 -9075
  46. package/src/analytics.js +393 -9
  47. package/src/app-context.js +211 -0
  48. package/src/app-stage.jsx +25 -24
  49. package/src/ardy/auto-fix-panel.css +108 -0
  50. package/src/ardy/collision-blockers.js +9 -3
  51. package/src/ardy/fix-collisions.js +10 -66
  52. package/src/ardy/ground.js +4 -4
  53. package/src/ardy/ik-drag.js +101 -0
  54. package/src/ardy/ik-key-json.js +43 -0
  55. package/src/ardy/ik.js +122 -76
  56. package/src/ardy/physics-panel.css +2 -0
  57. package/src/ardy/physics-panel.jsx +17 -5
  58. package/src/ardy/platform-fit-panel.css +64 -0
  59. package/src/ardy/platform-fit-panel.jsx +40 -0
  60. package/src/ardy/platform-fit.js +326 -0
  61. package/src/ardy/playback.js +166 -20
  62. package/src/ardy/range-pin.js +299 -0
  63. package/src/ardy/timeline-coordinates.js +26 -0
  64. package/src/ardy/timeline.css +1199 -0
  65. package/src/ardy/timeline.jsx +149 -41
  66. package/src/ardy/waypoints.js +2 -2
  67. package/src/asset-pane.css +462 -0
  68. package/src/asset-pane.jsx +461 -232
  69. package/src/command-bus.js +349 -0
  70. package/src/commands/ai.js +31 -0
  71. package/src/commands/cast.js +140 -0
  72. package/src/commands/elements/character.js +13 -0
  73. package/src/commands/elements/motion.js +14 -0
  74. package/src/commands/elements/object.js +22 -0
  75. package/src/commands/elements/scene.js +16 -0
  76. package/src/commands/elements/shot.js +9 -0
  77. package/src/commands/elements/stage.js +11 -0
  78. package/src/commands/elements.js +155 -0
  79. package/src/commands/export.js +20 -0
  80. package/src/commands/index.js +40 -0
  81. package/src/commands/motion.js +231 -0
  82. package/src/commands/objects.js +174 -0
  83. package/src/commands/project.js +53 -0
  84. package/src/commands/scene.js +81 -0
  85. package/src/commands/shared.js +19 -0
  86. package/src/commands/shot.js +103 -0
  87. package/src/commands/stage.js +17 -0
  88. package/src/commands/view.js +42 -0
  89. package/src/document-store.js +144 -0
  90. package/src/domains/cast.js +1007 -0
  91. package/src/domains/motion.js +3219 -0
  92. package/src/domains/objects.js +932 -0
  93. package/src/domains/scenes.js +817 -0
  94. package/src/domains/shots.js +404 -0
  95. package/src/domains/stage.js +93 -0
  96. package/src/dualview.jsx +113 -57
  97. package/src/facing-marks.js +3 -0
  98. package/src/first-success-guide.jsx +14 -14
  99. package/src/grid-view.js +18 -5
  100. package/src/hierarchy-model.js +24 -16
  101. package/src/hierarchy-panel.css +386 -0
  102. package/src/hierarchy-panel.jsx +150 -21
  103. package/src/{fal-motion-client.js → i2v-motion-client.js} +13 -13
  104. package/src/i2v-motion-studio.jsx +180 -0
  105. package/src/ik-camera.js +3 -0
  106. package/src/main.jsx +6 -0
  107. package/src/motion/generation.js +105 -0
  108. package/src/motion-readiness-ui.jsx +4 -4
  109. package/src/motion-readiness.js +11 -0
  110. package/src/motion-trail.js +366 -9
  111. package/src/object-gizmo.jsx +7 -1
  112. package/src/otio.js +2 -1
  113. package/src/panels/CameraPanel.jsx +60 -0
  114. package/src/panels/CharacterTransformPanel.jsx +48 -0
  115. package/src/panels/EnvironmentPanel.jsx +40 -0
  116. package/src/panels/Foldout.jsx +32 -0
  117. package/src/panels/LightPanel.jsx +24 -0
  118. package/src/panels/ObjectTransformPanel.jsx +554 -0
  119. package/src/panels/PosePanel.jsx +97 -0
  120. package/src/panels/ProjectPanel.jsx +33 -0
  121. package/src/panels/PromptBlocksPanel.jsx +442 -0
  122. package/src/panels/PropsPanel.jsx +70 -0
  123. package/src/panels/ReferenceImageField.jsx +91 -0
  124. package/src/panels/RigControlPanel.jsx +78 -0
  125. package/src/panels/RigPanel.jsx +37 -0
  126. package/src/panels/SubjectBox.jsx +47 -0
  127. package/src/panels/SubjectsPanel.jsx +39 -0
  128. package/src/panels/VideoCapturePanel.jsx +180 -0
  129. package/src/panels/details.css +1171 -0
  130. package/src/panels/motion.css +197 -0
  131. package/src/panels/pose.css +39 -0
  132. package/src/planview.jsx +60 -31
  133. package/src/posestudio.jsx +59 -4
  134. package/src/project-browser.css +1073 -0
  135. package/src/project-browser.jsx +284 -123
  136. package/src/range-pin-object-transform.js +19 -0
  137. package/src/range-pin-panel.css +518 -0
  138. package/src/range-pin-panel.jsx +344 -0
  139. package/src/result-modal.jsx +5 -5
  140. package/src/room.jsx +79 -51
  141. package/src/scene-objects.js +37 -1
  142. package/src/scenes.js +5 -0
  143. package/src/semantic-edit.js +2 -0
  144. package/src/settings-menu.jsx +39 -105
  145. package/src/shell/BottomDock.jsx +450 -0
  146. package/src/shell/DetailsSlot.jsx +513 -0
  147. package/src/shell/LibrarySlot.jsx +86 -0
  148. package/src/shell/MenuBar.jsx +601 -0
  149. package/src/shell/OutlinerSlot.jsx +50 -0
  150. package/src/shell/PreferencesDialog.jsx +538 -0
  151. package/src/shell/PreferencesSlot.jsx +40 -0
  152. package/src/shell/StatusBar.jsx +125 -0
  153. package/src/shell/StudioShell.jsx +92 -0
  154. package/src/shell/TopBar.jsx +129 -0
  155. package/src/shell/ViewportToolbar.jsx +656 -0
  156. package/src/shell/agent-glass.css +296 -0
  157. package/src/shell/dock.css +93 -0
  158. package/src/shell/glass-regions.css +985 -0
  159. package/src/shell/glass.css +687 -0
  160. package/src/shell/log-store.js +52 -0
  161. package/src/shell/mode.css +9 -0
  162. package/src/shell/preferences.css +651 -0
  163. package/src/shell/shell.css +282 -0
  164. package/src/shell/studio-shell-context.js +19 -0
  165. package/src/shell/topbar.css +435 -0
  166. package/src/shell/viewport.css +451 -0
  167. package/src/store/authored-intent.js +22 -0
  168. package/src/store/runtime-adapters.js +92 -0
  169. package/src/store/scene-stage.js +21 -0
  170. package/src/store/use-document-store.js +14 -0
  171. package/src/studio-actions.js +210 -0
  172. package/src/studio-agent-commands.js +41 -271
  173. package/src/studio-agent-context.js +39 -10
  174. package/src/studio-agent-motion.js +122 -373
  175. package/src/studio-agent-protocol.js +67 -31
  176. package/src/studio-app-binding.js +379 -0
  177. package/src/studio-contact-sheet.js +73 -0
  178. package/src/studio-elements.js +58 -51
  179. package/src/styles/themes.css +200 -0
  180. package/src/styles/tokens.css +69 -0
  181. package/src/styles.css +460 -1742
  182. package/src/theme.js +44 -0
  183. package/src/timeline-extent.js +16 -0
  184. package/src/trail-key-conflicts.js +39 -0
  185. package/src/trail-pick.js +59 -0
  186. package/src/ui.jsx +8 -8
  187. package/src/use-case-question.jsx +66 -0
  188. package/src/workflow/AgentPanel.jsx +82 -69
  189. package/src/workflow/agent-client.js +12 -2
  190. package/src/workflow/agent-panel.css +34 -32
  191. package/src/workflow/cozy-scene-node.css +2 -0
  192. package/src/workflow/workflow.css +2 -0
  193. package/tools/ardy/__pycache__/cclay_gvhmr_worker.cpython-313.pyc +0 -0
  194. package/tools/ardy/bridge.mjs +164 -146
  195. package/tools/ardy/visual-qa.mjs +5 -5
  196. package/tools/bench/EXP3.md +61 -0
  197. package/tools/bench/cclay_bench_extract_incam.py +125 -0
  198. package/tools/bench/cclay_bench_extract_obs.py +204 -0
  199. package/tools/bench/cclay_bench_runner.py +42 -0
  200. package/tools/bench/cube-contact.mjs +162 -0
  201. package/tools/bench/exp3.mjs +152 -0
  202. package/tools/bench/extract-bench-lib.mjs +216 -0
  203. package/tools/bench/extract-bench.mjs +302 -0
  204. package/tools/bench/fal-generate.mjs +88 -0
  205. package/tools/bench/fit/README.md +189 -0
  206. package/tools/bench/fit/camera.mjs +59 -0
  207. package/tools/bench/fit/contact.mjs +419 -0
  208. package/tools/bench/fit/footlock.mjs +234 -0
  209. package/tools/bench/fit/motion.mjs +85 -0
  210. package/tools/bench/fit/pin.mjs +27 -0
  211. package/tools/bench/fit/remote.mjs +105 -0
  212. package/tools/bench/fit-bench.mjs +126 -0
  213. package/tools/bench/fit-sanity.mjs +45 -0
  214. package/tools/bench/metrics.mjs +313 -0
  215. package/tools/bench/obs/depth.mjs +258 -0
  216. package/tools/bench/obs/extrinsics.mjs +175 -0
  217. package/tools/bench/obs/fit_mannequin_betas.py +340 -0
  218. package/tools/bench/obs/ground.mjs +294 -0
  219. package/tools/bench/obs/heading.mjs +151 -0
  220. package/tools/bench/obs/ladder.mjs +555 -0
  221. package/tools/bench/obs/mannequin-betas.json +124 -0
  222. package/tools/bench/obs/remote.mjs +83 -0
  223. package/tools/bench/obs/rest_joints.py +52 -0
  224. package/tools/bench/obs/ybot-targets.mjs +49 -0
  225. package/tools/bench/obs-bench.mjs +520 -0
  226. package/tools/bench/score.mjs +343 -0
  227. package/tools/bench/summarize.mjs +86 -0
  228. package/tools/dev/pages/privacy.html +13 -8
  229. package/tools/gt-render/browser.mjs +159 -0
  230. package/tools/gt-render/camera-math.mjs +276 -0
  231. package/tools/gt-render/page.mjs +267 -0
  232. package/tools/gt-render/render.mjs +514 -0
  233. package/tools/gt-render/scene-box.mjs +31 -0
  234. package/tools/gt-render/take-transform.mjs +87 -0
  235. package/tools/morphgs/assets/gen_truth.py +39 -0
  236. package/tools/morphgs/assets/mesh_ori_rig.txt +27 -0
  237. package/tools/morphgs/assets/playback-check.mjs +30 -0
  238. package/tools/morphgs/demo-gate.sh +26 -0
  239. package/tools/morphgs/fbx2morphgs.mjs +68 -0
  240. package/tools/morphgs/morphgs-to-cskel27.mjs +105 -0
  241. package/tools/morphgs/patches/preprocess_src-none-mode.patch +76 -0
  242. package/tools/morphgs/setup-on-cluster.sh +126 -0
  243. package/tools/qa/css-rule-usage.mjs +283 -0
  244. package/tools/qa/studio-control-count.mjs +70 -32
  245. package/tools/run-tests.mjs +134 -12
  246. package/tools/track/DESIGN.md +513 -0
  247. package/tools/track/backfill-provenance.mjs +116 -0
  248. package/tools/track/budget.mjs +8 -0
  249. package/tools/track/check-rig.mjs +209 -0
  250. package/tools/track/diagnostics.schema.json +60 -0
  251. package/tools/track/export-rig.mjs +256 -0
  252. package/tools/track/fallback.mjs +7 -0
  253. package/tools/track/fk-parity-fixture.mjs +161 -0
  254. package/tools/track/gate.mjs +349 -0
  255. package/tools/track/masks.mjs +318 -0
  256. package/tools/track/metrics.mjs +201 -0
  257. package/tools/track/publish-obs.mjs +108 -0
  258. package/tools/track/py/check_env.py +61 -0
  259. package/tools/track/py/eval_lr.py +229 -0
  260. package/tools/track/py/lr_viterbi.py +294 -0
  261. package/tools/track/py/masks.py +405 -0
  262. package/tools/track/py/objective.py +382 -0
  263. package/tools/track/py/rig.py +508 -0
  264. package/tools/track/py/scene.py +326 -0
  265. package/tools/track/py/test_joint_indices.py +123 -0
  266. package/tools/track/py/test_lr_viterbi.py +195 -0
  267. package/tools/track/py/test_masks.py +229 -0
  268. package/tools/track/py/test_rig.py +288 -0
  269. package/tools/track/py/test_scene.py +354 -0
  270. package/tools/track/py/test_track.py +328 -0
  271. package/tools/track/py/track.py +411 -0
  272. package/tools/track/remote.mjs +140 -0
  273. package/tools/track/rig-dump.mjs +227 -0
  274. package/tools/track/run-box-tests.mjs +14 -0
  275. package/tools/track/run-box.mjs +168 -0
  276. package/tools/track/setup-box.sh +59 -0
  277. package/tools/track/study-2d.mjs +728 -0
  278. package/dist/assets/analytics-B1hnH66c.js +0 -1
  279. package/dist/assets/app-BWusbqgO.js +0 -4861
  280. package/dist/assets/app-qKDo4PBX.css +0 -1
  281. package/dist/assets/shot-prompt-CHUtw6af.js +0 -17
  282. package/dist/assets/shot-prompt-C_g2BHVc.css +0 -1
  283. package/dist/assets/workflow-CVwzMhz2.js +0 -179
  284. package/dist/assets/workflow-DsudxKHF.css +0 -1
  285. package/src/fal-motion-studio.jsx +0 -180
  286. package/src/scene-history.js +0 -129
@@ -0,0 +1,294 @@
1
+ const FOOT_NAMES = ["L_ankle", "R_ankle", "L_foot", "R_foot"];
2
+ const ANKLE_INDEX = { L_ankle: 15, R_ankle: 16 };
3
+ const LOGIT_INDEX = { L_ankle: 0, L_foot: 1, R_ankle: 2, R_foot: 3 };
4
+ const OPPOSITE = { L_ankle: "R_ankle", R_ankle: "L_ankle" };
5
+
6
+ function valuesOf(value) {
7
+ return value?.data ?? value;
8
+ }
9
+
10
+ function flattenMatrix(matrix) {
11
+ return Array.isArray(matrix?.[0]) ? matrix.flat() : Array.from(matrix ?? []);
12
+ }
13
+
14
+ function matrix3(matrix, name) {
15
+ const flat = flattenMatrix(matrix);
16
+ if (flat.length !== 9 || flat.some((value) => !Number.isFinite(Number(value)))) throw new Error(`${name}: expected a finite 3x3 matrix`);
17
+ return flat.map(Number);
18
+ }
19
+
20
+ function vector3(value, name) {
21
+ const data = valuesOf(value);
22
+ if (!data || data.length !== 3 || Array.from(data).some((item) => !Number.isFinite(Number(item)))) throw new Error(`${name}: expected a finite 3-vector`);
23
+ return Array.from(data, Number);
24
+ }
25
+
26
+ function frameVectors(value, frames, name) {
27
+ const data = valuesOf(value);
28
+ if (Array.isArray(data) && data.length === frames && data.every((row) => Array.isArray(row))) {
29
+ return data.map((row, frame) => {
30
+ if (row.length !== 3 || row.some((item) => !Number.isFinite(Number(item)))) throw new Error(`${name}[${frame}]: expected a finite 3-vector`);
31
+ return row.map(Number);
32
+ });
33
+ }
34
+ if (!data || data.length !== frames * 3) throw new Error(`${name}: expected [${frames}, 3]`);
35
+ return Array.from({ length: frames }, (_, frame) => Array.from(data.slice(frame * 3, frame * 3 + 3), Number));
36
+ }
37
+
38
+ function frameKeypoints(value, frames) {
39
+ const data = valuesOf(value);
40
+ if (Array.isArray(data) && data.length === frames && data.every((row) => Array.isArray(row))) {
41
+ return data.map((row, frame) => {
42
+ if (row.length < 17 || row.some((point) => !Array.isArray(point) || point.length < 3)) throw new Error(`kp2d[${frame}]: expected at least 17 [x, y, confidence] points`);
43
+ return row.map((point) => point.slice(0, 3).map(Number));
44
+ });
45
+ }
46
+ if (!data || data.length !== frames * 17 * 3) throw new Error(`kp2d: expected [${frames}, 17, 3]`);
47
+ return Array.from({ length: frames }, (_, frame) => Array.from({ length: 17 }, (_, joint) => {
48
+ const offset = (frame * 17 + joint) * 3;
49
+ return Array.from(data.slice(offset, offset + 3), Number);
50
+ }));
51
+ }
52
+
53
+ function frameLogits(value, frames) {
54
+ const data = valuesOf(value);
55
+ if (Array.isArray(data) && data.length === frames && data.every((row) => Array.isArray(row))) {
56
+ return data.map((row, frame) => {
57
+ if (row.length < 6) throw new Error(`static_conf_logits[${frame}]: expected 6 logits`);
58
+ return row.slice(0, 6).map(Number);
59
+ });
60
+ }
61
+ if (!data || data.length !== frames * 6) throw new Error(`static_conf_logits: expected [${frames}, 6]`);
62
+ return Array.from({ length: frames }, (_, frame) => Array.from(data.slice(frame * 6, frame * 6 + 6), Number));
63
+ }
64
+
65
+ function footPoint(footW, frame, name) {
66
+ if (Array.isArray(footW)) {
67
+ const row = footW[frame];
68
+ if (Array.isArray(row)) {
69
+ const index = FOOT_NAMES.indexOf(name);
70
+ return row[index];
71
+ }
72
+ return row?.[name];
73
+ }
74
+ return footW?.[name]?.[frame];
75
+ }
76
+
77
+ function normalizedFootFrames(footW, frames) {
78
+ return Array.from({ length: frames }, (_, frame) => Object.fromEntries(FOOT_NAMES.map((name) => {
79
+ const point = footPoint(footW, frame, name);
80
+ return [name, point && point.length === 3 ? Array.from(point, Number) : null];
81
+ })));
82
+ }
83
+
84
+ function normalizeCamera(camera) {
85
+ if (!camera) throw new Error("camera: expected K, R_c2w, and t_c2w");
86
+ const K = matrix3(camera.K ?? [[camera.fx, 0, camera.cx], [0, camera.fy, camera.cy], [0, 0, 1]], "camera.K");
87
+ let rotation = camera.R_c2w ?? camera.Rc2w;
88
+ let translation = camera.t_c2w ?? camera.tc2w;
89
+ if (!rotation && camera.worldToCamera) {
90
+ const worldToCamera = matrix3(camera.worldToCamera, "camera.worldToCamera");
91
+ rotation = [worldToCamera[0], worldToCamera[3], worldToCamera[6], worldToCamera[1], worldToCamera[4], worldToCamera[7], worldToCamera[2], worldToCamera[5], worldToCamera[8]];
92
+ const tW2C = [worldToCamera[3], worldToCamera[7], worldToCamera[11]];
93
+ translation = [
94
+ -(rotation[0] * tW2C[0] + rotation[1] * tW2C[1] + rotation[2] * tW2C[2]),
95
+ -(rotation[3] * tW2C[0] + rotation[4] * tW2C[1] + rotation[5] * tW2C[2]),
96
+ -(rotation[6] * tW2C[0] + rotation[7] * tW2C[1] + rotation[8] * tW2C[2]),
97
+ ];
98
+ }
99
+ return { K, R_c2w: matrix3(rotation, "camera.R_c2w"), t_c2w: vector3(translation, "camera.t_c2w") };
100
+ }
101
+
102
+ function pixelRay(u, v, camera) {
103
+ const { K, R_c2w, t_c2w } = camera;
104
+ const cameraDirection = [(u - K[2]) / K[0], (v - K[5]) / K[4], 1];
105
+ const direction = [
106
+ R_c2w[0] * cameraDirection[0] + R_c2w[1] * cameraDirection[1] + R_c2w[2] * cameraDirection[2],
107
+ R_c2w[3] * cameraDirection[0] + R_c2w[4] * cameraDirection[1] + R_c2w[5] * cameraDirection[2],
108
+ R_c2w[6] * cameraDirection[0] + R_c2w[7] * cameraDirection[1] + R_c2w[8] * cameraDirection[2],
109
+ ];
110
+ const length = Math.hypot(...direction);
111
+ return { origin: t_c2w, direction: direction.map((item) => item / length) };
112
+ }
113
+
114
+ function median(values) {
115
+ if (!values.length) return 0;
116
+ const sorted = values.slice().sort((a, b) => a - b);
117
+ const middle = Math.floor(sorted.length / 2);
118
+ return sorted.length % 2 ? sorted[middle] : (sorted[middle - 1] + sorted[middle]) / 2;
119
+ }
120
+
121
+ function sigmoid(logit) {
122
+ if (!Number.isFinite(logit)) return 0;
123
+ if (logit >= 0) {
124
+ const e = Math.exp(-logit);
125
+ return 1 / (1 + e);
126
+ }
127
+ const e = Math.exp(logit);
128
+ return e / (1 + e);
129
+ }
130
+
131
+ function translatePoint(point, correction) {
132
+ return [point[0] + correction[0], point[1], point[2] + correction[2]];
133
+ }
134
+
135
+ function shiftJoints(jointsW, corrections) {
136
+ if (!Array.isArray(jointsW)) return jointsW;
137
+ return jointsW.map((frame, index) => Array.isArray(frame) ? frame.map((point) => point?.length === 3 ? translatePoint(point, corrections[index]) : point) : frame);
138
+ }
139
+
140
+ function shiftFeet(footW, corrections, frames) {
141
+ if (Array.isArray(footW)) return footW.map((frame, index) => Array.isArray(frame) ? frame.map((point) => point?.length === 3 ? translatePoint(point, corrections[index]) : point) : Object.fromEntries(FOOT_NAMES.map((name) => [name, frame?.[name]?.length === 3 ? translatePoint(frame[name], corrections[index]) : frame?.[name]])));
142
+ if (!footW || typeof footW !== "object") return footW;
143
+ return Object.fromEntries(Object.entries(footW).map(([name, track]) => [name, Array.isArray(track) ? track.map((point, index) => point?.length === 3 ? translatePoint(point, corrections[index]) : point) : track]));
144
+ }
145
+
146
+ function interpolateCorrections(samples, frames) {
147
+ const corrections = Array.from({ length: frames }, () => [0, 0, 0]);
148
+ const contactMask = samples.map((items) => items.length > 0);
149
+ const runs = [];
150
+ for (let frame = 0; frame < frames;) {
151
+ if (!contactMask[frame]) {
152
+ frame += 1;
153
+ continue;
154
+ }
155
+ const start = frame;
156
+ while (frame < frames && contactMask[frame]) frame += 1;
157
+ const end = frame - 1;
158
+ const anchor = [0, 0, 0];
159
+ for (let axis of [0, 2]) anchor[axis] = median(samples.slice(start, end + 1).flatMap((items) => items.map((sample) => sample[axis])));
160
+ runs.push({ start, end, anchor, samples: samples.slice(start, end + 1).flat().length });
161
+ for (let at = start; at <= end; at += 1) corrections[at] = anchor.slice();
162
+ }
163
+ if (!runs.length) return { corrections, runs };
164
+ for (let index = 0; index + 1 < runs.length; index += 1) {
165
+ const left = runs[index], right = runs[index + 1];
166
+ const span = right.start - left.end;
167
+ for (let frame = left.end + 1; frame < right.start; frame += 1) {
168
+ const weight = (frame - left.end) / span;
169
+ corrections[frame] = [
170
+ left.anchor[0] + (right.anchor[0] - left.anchor[0]) * weight,
171
+ 0,
172
+ left.anchor[2] + (right.anchor[2] - left.anchor[2]) * weight,
173
+ ];
174
+ }
175
+ }
176
+ for (let frame = 0; frame < runs[0].start; frame += 1) corrections[frame] = runs[0].anchor.slice();
177
+ const last = runs[runs.length - 1];
178
+ for (let frame = last.end + 1; frame < frames; frame += 1) corrections[frame] = last.anchor.slice();
179
+ return { corrections, runs };
180
+ }
181
+
182
+ function footSpeed(feet, name, frame, frames, fps) {
183
+ const before = feet[Math.max(0, frame - 1)][name], after = feet[Math.min(frames - 1, frame + 1)][name];
184
+ const span = (Math.min(frames - 1, frame + 1) - Math.max(0, frame - 1)) / fps;
185
+ return before?.every(Number.isFinite) && after?.every(Number.isFinite) && span > 0 ? Math.hypot(after[0] - before[0], after[2] - before[2]) / span : Infinity;
186
+ }
187
+
188
+ /**
189
+ * Correct a world trajectory using 2D ankle observations at static floor
190
+ * contacts. Corrections are XZ translations represented as [dx, 0, dz].
191
+ * Optional `jointsW` and `footW` inputs receive the same rigid translation.
192
+ */
193
+ export function correctTrajectory({
194
+ rootW,
195
+ footW,
196
+ jointsW,
197
+ kp2d,
198
+ static_conf_logits: staticConfLogits,
199
+ camera,
200
+ ankleHeight,
201
+ contactProbability = 0.8,
202
+ maxFootHeight = 0.12,
203
+ minKeypointConfidence = 0.3,
204
+ // "kinematic" is the default. "hybrid": stance when EITHER GVHMR's static-contact logit OR the
205
+ // kinematic test says so. "kinematic" (default): lower foot, near its own low height,
206
+ // nearly still (absolutely, or relative to the other foot). GVHMR's logits are
207
+ // weak on the part-coloured mannequin (walk: right foot never exceeds 0.8), so
208
+ // kinematics carry real clips; logits carry clips whose planted foot still
209
+ // slides in the estimate. "logits": logits only.
210
+ stanceMode = "kinematic",
211
+ fps = 24,
212
+ stanceHeightMargin = 0.04,
213
+ maxStanceSpeed = 0.25,
214
+ minStanceFrames = 3,
215
+ } = {}) {
216
+ if (!Array.isArray(rootW) || !rootW.length) throw new Error("rootW: expected at least one frame");
217
+ if (!Number.isFinite(ankleHeight)) throw new Error("ankleHeight: expected a finite number");
218
+ if (!(contactProbability > 0 && contactProbability < 1)) throw new Error("contactProbability: expected a value in (0, 1)");
219
+ if (!(maxFootHeight > 0) || !Number.isFinite(maxFootHeight)) throw new Error("maxFootHeight: expected a positive number");
220
+ if (!(minKeypointConfidence >= 0) || !Number.isFinite(minKeypointConfidence)) throw new Error("minKeypointConfidence: expected a nonnegative number");
221
+ const frames = rootW.length;
222
+ const roots = frameVectors(rootW, frames, "rootW");
223
+ const feet = normalizedFootFrames(footW, frames);
224
+ const keypoints = frameKeypoints(kp2d, frames);
225
+ const logits = frameLogits(staticConfLogits, frames);
226
+ const intrinsics = normalizeCamera(camera);
227
+ const samples = Array.from({ length: frames }, () => []);
228
+ const contactByFoot = Object.fromEntries(Object.keys(OPPOSITE).map((name) => [name, []]));
229
+ if (!["hybrid", "kinematic", "logits"].includes(stanceMode)) throw new Error(`stanceMode: expected "hybrid", "kinematic" or "logits", got ${stanceMode}`);
230
+ const stance = Object.fromEntries(Object.keys(OPPOSITE).map((name) => {
231
+ const heights = feet.map((row) => row[name]?.[1]).filter(Number.isFinite).sort((a, b) => a - b);
232
+ const low = heights.length ? heights[Math.floor(heights.length * 0.1)] : NaN;
233
+ const flags = Array.from({ length: frames }, (_, frame) => {
234
+ const point = feet[frame][name];
235
+ const opposite = feet[frame][OPPOSITE[name]];
236
+ const keypoint = keypoints[frame][ANKLE_INDEX[name]];
237
+ const base = point?.every(Number.isFinite) && opposite?.every(Number.isFinite) && keypoint?.every(Number.isFinite) && keypoint[2] >= minKeypointConfidence && point[1] <= opposite[1];
238
+ if (!base) return false;
239
+ const logitOffset = LOGIT_INDEX[name];
240
+ const staticProbability = Math.max(sigmoid(logits[frame][logitOffset]), sigmoid(logits[frame][logitOffset + 1]));
241
+ const byLogits = staticProbability > contactProbability && point[1] < maxFootHeight;
242
+ if (stanceMode === "logits") return byLogits;
243
+ if (stanceMode === "hybrid" && byLogits) return true;
244
+ // The planted foot is the slower of the two, or absolutely still. A
245
+ // relative test survives the very error G3 corrects: a mis-scaled
246
+ // trajectory makes even the stance foot slide in world space.
247
+ const speed = footSpeed(feet, name, frame, frames, fps);
248
+ const otherSpeed = footSpeed(feet, OPPOSITE[name], frame, frames, fps);
249
+ return point[1] <= low + stanceHeightMargin && (speed < maxStanceSpeed || speed < 0.5 * otherSpeed);
250
+ });
251
+ // Short flickers are not a planted foot: keep runs of at least minStanceFrames.
252
+ for (let start = 0; start < frames;) {
253
+ if (!flags[start]) { start += 1; continue; }
254
+ let end = start;
255
+ while (end + 1 < frames && flags[end + 1]) end += 1;
256
+ if (end - start + 1 < minStanceFrames) for (let i = start; i <= end; i += 1) flags[i] = false;
257
+ start = end + 1;
258
+ }
259
+ return [name, flags];
260
+ }));
261
+ for (let frame = 0; frame < frames; frame += 1) {
262
+ for (const name of Object.keys(OPPOSITE)) {
263
+ const point = feet[frame][name];
264
+ const keypoint = keypoints[frame][ANKLE_INDEX[name]];
265
+ if (!stance[name][frame]) continue;
266
+ const ray = pixelRay(keypoint[0], keypoint[1], intrinsics);
267
+ if (Math.abs(ray.direction[1]) < 1e-10) continue;
268
+ const distance = (ankleHeight - ray.origin[1]) / ray.direction[1];
269
+ if (!(distance > 0) || !Number.isFinite(distance)) continue;
270
+ const observed = [ray.origin[0] + distance * ray.direction[0], ankleHeight, ray.origin[2] + distance * ray.direction[2]];
271
+ samples[frame].push([observed[0] - point[0], 0, observed[2] - point[2]]);
272
+ contactByFoot[name].push(frame);
273
+ }
274
+ }
275
+ const { corrections, runs } = interpolateCorrections(samples, frames);
276
+ const correctedRoot = roots.map((point, frame) => translatePoint(point, corrections[frame]));
277
+ const result = {
278
+ rootW: correctedRoot,
279
+ corrections,
280
+ contactFrames: samples.flatMap((items, frame) => items.length ? [frame] : []),
281
+ diagnostics: {
282
+ contactByFoot,
283
+ stanceRuns: runs.map((run) => ({ ...run, anchor: run.anchor.slice() })),
284
+ contactCount: samples.reduce((total, items) => total + items.length, 0),
285
+ interpolatedFrames: corrections.reduce((total, correction, frame) => total + (samples[frame].length ? 0 : (runs.length ? 1 : 0)), 0),
286
+ maxCorrectionM: Math.max(...corrections.map((correction) => Math.hypot(correction[0], correction[2]))),
287
+ planeY: ankleHeight,
288
+ thresholds: { stanceMode, contactProbability, maxFootHeight, minKeypointConfidence, stanceHeightMargin, maxStanceSpeed, minStanceFrames, fps },
289
+ },
290
+ };
291
+ if (jointsW !== undefined) result.jointsW = shiftJoints(jointsW, corrections);
292
+ if (footW !== undefined) result.footW = shiftFeet(footW, corrections, frames);
293
+ return result;
294
+ }
@@ -0,0 +1,151 @@
1
+ /**
2
+ * heading.mjs — remove GVHMR's short front/back yaw-flip runs from a SMPL
3
+ * global_orient series before temporal smoothing.
4
+ *
5
+ * On the faceless mannequin ViTPose swaps left/right for 1-8 frames, so the
6
+ * raw global_orient turns 97-178 deg about world vertical in one frame (or two
7
+ * or three) and turns back later. body_pose and pelvis do not flip. Gaussian
8
+ * smoothing then smears each such run into a visible spin, so the runs are
9
+ * removed here first.
10
+ *
11
+ * Mechanism:
12
+ * 1. Heading = twist of global_orient about world +Y (swing-twist
13
+ * decomposition). Pre-multiplying a world Ry(a) shifts it by exactly a
14
+ * and leaves the swing (tilt) unchanged.
15
+ * 2. Jump events = maximal clusters of consecutive frames whose yaw step
16
+ * exceeds minJumpDeg/2, kept when the cluster's net yaw exceeds
17
+ * minJumpDeg. A flip spread over 2-3 transitional frames is one event.
18
+ * 3. Paired runs: an entry event followed within maxRunFrames by the events
19
+ * whose summed yaw returns within minJumpDeg of zero. The run is every
20
+ * frame from the first frame of the entry event up to the frame before
21
+ * the last frame of the exit event.
22
+ * 4. Each run frame t is corrected by Ry(-offset(t)). offset(t) is the
23
+ * summed measured jump yaw up to t, not a fixed 180. Any exit mismatch r
24
+ * (flips that are not exactly opposite) is spread linearly over the
25
+ * run, so both joins step by only r/(L+1).
26
+ * 5. An unpaired event within maxRunFrames of the clip start or end marks
27
+ * a run touching frame 0 or the last frame. It is corrected toward the
28
+ * stable side when that side is longer than the run.
29
+ * 6. A flip spread over several frames leaves mid-flip frames that match
30
+ * neither heading. Their tilt is corrupted too (walk f24 tilts 33 deg
31
+ * against 8-12 deg neighbours). These frames only (event frames before
32
+ * the event's last frame, inside a run) are replaced by a slerp between
33
+ * their corrected neighbours. Every other run frame gets the yaw-only
34
+ * correction and keeps its tilt.
35
+ * Frames outside runs are returned as copies of their input values.
36
+ */
37
+
38
+ import { axisAngleToMatrix, matrixToAxisAngle } from "./extrinsics.mjs";
39
+
40
+ const DEG = 180 / Math.PI;
41
+
42
+ /** Wrap degrees into [-180, 180). */
43
+ const wrapDeg = (a) => a - 360 * Math.floor((a + 180) / 360);
44
+
45
+ /** Twist of an axis-angle rotation about world +Y, degrees. */
46
+ export function worldYawDeg(aa) {
47
+ const theta = Math.hypot(aa[0], aa[1], aa[2]);
48
+ const qw = Math.cos(theta / 2);
49
+ const qy = theta > 1e-12 ? (aa[1] / theta) * Math.sin(theta / 2) : aa[1] / 2;
50
+ return wrapDeg(2 * Math.atan2(qy, qw) * DEG);
51
+ }
52
+
53
+ function mul3(a, b) {
54
+ const o = new Array(9);
55
+ for (let i = 0; i < 3; i++) for (let j = 0; j < 3; j++) o[i * 3 + j] = a[i * 3] * b[j] + a[i * 3 + 1] * b[3 + j] + a[i * 3 + 2] * b[6 + j];
56
+ return o;
57
+ }
58
+
59
+ function slerpAA(a, b, f) {
60
+ const A = axisAngleToMatrix(a);
61
+ const At = [A[0], A[3], A[6], A[1], A[4], A[7], A[2], A[5], A[8]];
62
+ const rel = matrixToAxisAngle(mul3(At, axisAngleToMatrix(b)));
63
+ return matrixToAxisAngle(mul3(A, axisAngleToMatrix(rel.map((v) => v * f))));
64
+ }
65
+
66
+ function preRotateY(aa, deg) {
67
+ const a = deg / DEG, c = Math.cos(a), s = Math.sin(a);
68
+ const m = axisAngleToMatrix(Array.from(aa));
69
+ // Ry(a) * m, Ry = [c 0 s; 0 1 0; -s 0 c]
70
+ return matrixToAxisAngle([
71
+ c * m[0] + s * m[6], c * m[1] + s * m[7], c * m[2] + s * m[8],
72
+ m[3], m[4], m[5],
73
+ -s * m[0] + c * m[6], -s * m[1] + c * m[7], -s * m[2] + c * m[8],
74
+ ]);
75
+ }
76
+
77
+ /**
78
+ * @param {ArrayLike<number>[]} orientAA [T] axis-angle global_orient (world, +Y up)
79
+ * @param {{ maxRunFrames?: number, minJumpDeg?: number }} [options]
80
+ * @returns {{ orient: number[][], runs: { start: number, end: number, yawDeg: number }[] }}
81
+ */
82
+ export function unwrapHeadingFlips(orientAA, { maxRunFrames = 12, minJumpDeg = 60 } = {}) {
83
+ const T = orientAA.length;
84
+ const yaw = Array.from(orientAA, worldYawDeg);
85
+ const d = yaw.map((y, t) => (t ? wrapDeg(y - yaw[t - 1]) : 0));
86
+
87
+ const events = [];
88
+ for (let t = 1; t < T;) {
89
+ if (!(Math.abs(d[t]) > minJumpDeg / 2)) { t++; continue; }
90
+ let e = t, sum = 0;
91
+ while (e < T && Math.abs(d[e]) > minJumpDeg / 2) sum += d[e++];
92
+ if (Math.abs(wrapDeg(sum)) > minJumpDeg) events.push({ first: t, last: e - 1, sum: wrapDeg(sum), used: false });
93
+ t = e;
94
+ }
95
+ // Summed wrapped steps over every event frame in [from, to].
96
+ const eventYaw = (from, to) => {
97
+ let s = 0;
98
+ for (const ev of events) for (let k = Math.max(ev.first, from); k <= Math.min(ev.last, to); k++) s += d[k];
99
+ return s;
100
+ };
101
+
102
+ // runs: { start, end, yawDeg, offset(t) }
103
+ let runs = [];
104
+ outer: for (let i = 0; i < events.length; i++) {
105
+ let net = events[i].sum;
106
+ for (let j = i + 1; j < events.length; j++) {
107
+ const start = events[i].first, end = events[j].last - 1;
108
+ if (end - start + 1 > maxRunFrames) break;
109
+ net += events[j].sum;
110
+ if (Math.abs(wrapDeg(net)) < minJumpDeg) {
111
+ const L = end - start + 1, r = wrapDeg(eventYaw(start, events[j].last));
112
+ runs.push({ start, end, yawDeg: events[i].sum, offset: (t) => eventYaw(start, t) - (r * (t - start + 1)) / (L + 1) });
113
+ for (let k = i; k <= j; k++) events[k].used = true;
114
+ i = j;
115
+ continue outer;
116
+ }
117
+ }
118
+ }
119
+
120
+ const unpaired = events.filter((ev) => !ev.used);
121
+ const head = unpaired.filter((ev) => ev.last <= maxRunFrames).pop();
122
+ if (head) {
123
+ const next = events.find((ev) => ev.first > head.last);
124
+ if ((next ? next.first : T) - head.last > head.last) {
125
+ const end = head.last - 1;
126
+ runs = runs.filter((run) => run.start > end);
127
+ runs.unshift({ start: 0, end, yawDeg: wrapDeg(-eventYaw(0, head.last)), offset: (t) => -eventYaw(t + 1, head.last) });
128
+ head.used = true;
129
+ }
130
+ }
131
+ const tail = unpaired.find((ev) => !ev.used && T - ev.first <= maxRunFrames);
132
+ if (tail) {
133
+ const prev = events.filter((ev) => ev.last < tail.first).pop();
134
+ const start = tail.first;
135
+ if (start - (prev ? prev.last : 0) > T - start) {
136
+ runs = runs.filter((run) => run.end < start);
137
+ runs.push({ start, end: T - 1, yawDeg: wrapDeg(eventYaw(start, T - 1)), offset: (t) => eventYaw(start, t) });
138
+ }
139
+ }
140
+
141
+ const orient = Array.from(orientAA, (aa) => Array.from(aa));
142
+ for (const run of runs) for (let t = run.start; t <= run.end; t++) orient[t] = preRotateY(orientAA[t], -run.offset(t));
143
+ for (const run of runs) {
144
+ for (const ev of events) {
145
+ if (ev.last - ev.first < 1 || ev.first > run.end + 1 || ev.last < run.start) continue;
146
+ const lo = ev.first - 1, hi = ev.last;
147
+ for (let k = Math.max(ev.first, run.start); k <= Math.min(ev.last - 1, run.end); k++) orient[k] = slerpAA(orient[lo], orient[hi], (k - lo) / (hi - lo));
148
+ }
149
+ }
150
+ return { orient, runs: runs.map(({ start, end, yawDeg }) => ({ start, end, yawDeg: Math.round(yawDeg * 10) / 10 })) };
151
+ }