VideoInference / model.json
Zihanj's picture
Upload curated v1 multihead ONNX bundle
21d687a verified
Raw
History Blame Contribute Delete
4.9 kB
{
"model_file": "action_classifier_curated_v1_multihead_w10_cuda.onnx",
"window": 10,
"feat_per_frame": 731,
"input": {
"name": "skeleton_window",
"shape": [
"batch",
10,
731
],
"dtype": "float32",
"description": "Sliding window of skeleton features. Each frame is a 731-dim vector: hip-centred (x,y,z,conf)×73 + motion_mag + ref_offset(dx,dy,dz)×73 + ref_std×73. World coordinates (metres, hip-centred). See docs/arch/v1.1/README.md for feature layout."
},
"outputs": [
{
"name": "locomotion",
"shape": [
"batch",
18
],
"dtype": "float32",
"classes": [
"walk",
"fastwalk",
"run",
"lowrun",
"lstride",
"lstride_back",
"rstride",
"rstride_back",
"lfwdstride",
"lfwdstride_back",
"rfwdstride",
"rfwdstride_back",
"lbkstep",
"lbkstep_back",
"rbkstep",
"rbkstep_back",
"highknee",
"noaction"
],
"label_map": {
"walk": 0,
"fastwalk": 1,
"run": 2,
"lowrun": 3,
"lstride": 4,
"lstride_back": 5,
"rstride": 6,
"rstride_back": 7,
"lfwdstride": 8,
"lfwdstride_back": 9,
"rfwdstride": 10,
"rfwdstride_back": 11,
"lbkstep": 12,
"lbkstep_back": 13,
"rbkstep": 14,
"rbkstep_back": 15,
"highknee": 16,
"noaction": 17
},
"noaction_index": 17
},
{
"name": "jump_squat",
"shape": [
"batch",
5
],
"dtype": "float32",
"classes": [
"jump",
"jump_back",
"squat",
"squat_back",
"noaction"
],
"label_map": {
"jump": 0,
"jump_back": 1,
"squat": 2,
"squat_back": 3,
"noaction": 4
},
"noaction_index": 4
},
{
"name": "arms",
"shape": [
"batch",
6
],
"dtype": "float32",
"classes": [
"larmchop",
"rarmchop",
"multichop",
"larmlowside",
"rarmlowside",
"noaction"
],
"label_map": {
"larmchop": 0,
"rarmchop": 1,
"multichop": 2,
"larmlowside": 3,
"rarmlowside": 4,
"noaction": 5
},
"noaction_index": 5
},
{
"name": "hands",
"shape": [
"batch",
7
],
"dtype": "float32",
"classes": [
"lhandgesture",
"rhandgesture",
"bothhandgesture",
"lhandchest",
"rhandchest",
"pickup",
"noaction"
],
"label_map": {
"lhandgesture": 0,
"rhandgesture": 1,
"bothhandgesture": 2,
"lhandchest": 3,
"rhandchest": 4,
"pickup": 5,
"noaction": 6
},
"noaction_index": 6
},
{
"name": "seals",
"shape": [
"batch",
6
],
"dtype": "float32",
"classes": [
"lhandseal",
"rhandseal",
"headseal",
"chestseal",
"bellyseal",
"noaction"
],
"label_map": {
"lhandseal": 0,
"rhandseal": 1,
"headseal": 2,
"chestseal": 3,
"bellyseal": 4,
"noaction": 5
},
"noaction_index": 5
},
{
"name": "head",
"shape": [
"batch",
5
],
"dtype": "float32",
"classes": [
"headdown",
"headup",
"lheadturn",
"rheadturn",
"noaction"
],
"label_map": {
"headdown": 0,
"headup": 1,
"lheadturn": 2,
"rheadturn": 3,
"noaction": 4
},
"noaction_index": 4
},
{
"name": "torso",
"shape": [
"batch",
5
],
"dtype": "float32",
"classes": [
"llean",
"rlean",
"lshoulderturn",
"rshoulderturn",
"noaction"
],
"label_map": {
"llean": 0,
"rlean": 1,
"lshoulderturn": 2,
"rshoulderturn": 3,
"noaction": 4
},
"noaction_index": 4
},
{
"name": "poses",
"shape": [
"batch",
5
],
"dtype": "float32",
"classes": [
"tpose",
"tpose_back",
"ypose",
"ypose_back",
"noaction"
],
"label_map": {
"tpose": 0,
"tpose_back": 1,
"ypose": 2,
"ypose_back": 3,
"noaction": 4
},
"noaction_index": 4
}
]
}