Files
hao-ai-lab-FastVideo/examples/distill/MatrixGame2.0/validation.json
T
2026-03-04 04:27:21 +00:00

324 lines
19 KiB
JSON

{
"data": [
{
"caption": "000009_Bravo_instance_000",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000009_Bravo_instance_000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000009_Bravo_instance_000_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000009_Bravo_instance_000_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000009_Bravo_instance_001",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000009_Bravo_instance_001.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000009_Bravo_instance_001_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000009_Bravo_instance_001_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000010_Alpha_instance_000",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000010_Alpha_instance_000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000010_Alpha_instance_000_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000010_Alpha_instance_000_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000010_Alpha_instance_001",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000010_Alpha_instance_001.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000010_Alpha_instance_001_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000010_Alpha_instance_001_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000010_Bravo_instance_000",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000010_Bravo_instance_000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000010_Bravo_instance_000_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000010_Bravo_instance_000_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000010_Bravo_instance_001",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000010_Bravo_instance_001.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000010_Bravo_instance_001_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000010_Bravo_instance_001_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000011_Alpha_instance_000",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000011_Alpha_instance_000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000011_Alpha_instance_000_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000011_Alpha_instance_000_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000011_Alpha_instance_001",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000011_Alpha_instance_001.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000011_Alpha_instance_001_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000011_Alpha_instance_001_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000011_Bravo_instance_000",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000011_Bravo_instance_000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000011_Bravo_instance_000_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000011_Bravo_instance_000_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000011_Bravo_instance_001",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000011_Bravo_instance_001.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000011_Bravo_instance_001_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000011_Bravo_instance_001_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000012_Alpha_instance_000",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000012_Alpha_instance_000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000012_Alpha_instance_000_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000012_Alpha_instance_000_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000012_Alpha_instance_001",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000012_Alpha_instance_001.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000012_Alpha_instance_001_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000012_Alpha_instance_001_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000012_Bravo_instance_000",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000012_Bravo_instance_000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000012_Bravo_instance_000_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/translationEval/test/000012_Bravo_instance_000_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000008_Bravo_instance_000",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000008_Bravo_instance_000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000008_Bravo_instance_000_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000008_Bravo_instance_000_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000008_Bravo_instance_001",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000008_Bravo_instance_001.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000008_Bravo_instance_001_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000008_Bravo_instance_001_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000009_Alpha_instance_000",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000009_Alpha_instance_000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000009_Alpha_instance_000_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000009_Alpha_instance_000_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000009_Alpha_instance_001",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000009_Alpha_instance_001.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000009_Alpha_instance_001_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000009_Alpha_instance_001_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000009_Bravo_instance_000",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000009_Bravo_instance_000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000009_Bravo_instance_000_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000009_Bravo_instance_000_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000009_Bravo_instance_001",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000009_Bravo_instance_001.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000009_Bravo_instance_001_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000009_Bravo_instance_001_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000010_Alpha_instance_000",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000010_Alpha_instance_000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000010_Alpha_instance_000_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000010_Alpha_instance_000_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000010_Alpha_instance_001",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000010_Alpha_instance_001.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000010_Alpha_instance_001_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000010_Alpha_instance_001_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000010_Bravo_instance_000",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000010_Bravo_instance_000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000010_Bravo_instance_000_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000010_Bravo_instance_000_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000010_Bravo_instance_001",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000010_Bravo_instance_001.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000010_Bravo_instance_001_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000010_Bravo_instance_001_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000010_Bravo_instance_001",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000010_Bravo_instance_001.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000010_Bravo_instance_001_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000010_Bravo_instance_001_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000011_Alpha_instance_000",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000011_Alpha_instance_000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000011_Alpha_instance_000_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000011_Alpha_instance_000_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000011_Alpha_instance_001",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000011_Alpha_instance_001.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000011_Alpha_instance_001_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000011_Alpha_instance_001_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000011_Bravo_instance_000",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000011_Bravo_instance_000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000011_Bravo_instance_000_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000011_Bravo_instance_000_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000011_Bravo_instance_001",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000011_Bravo_instance_001.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000011_Bravo_instance_001_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/oneLooksAwayEval/test/000011_Bravo_instance_001_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000015_Alpha_instance_000",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/structureEval/test/000015_Alpha_instance_000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/structureEval/test/000015_Alpha_instance_000_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/structureEval/test/000015_Alpha_instance_000_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000015_Alpha_instance_001",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/structureEval/test/000015_Alpha_instance_001.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/structureEval/test/000015_Alpha_instance_001_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/structureEval/test/000015_Alpha_instance_001_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000015_Bravo_instance_000",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/structureEval/test/000015_Bravo_instance_000.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/structureEval/test/000015_Bravo_instance_000_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/structureEval/test/000015_Bravo_instance_000_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
},
{
"caption": "000015_Bravo_instance_001",
"image_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/structureEval/test/000015_Bravo_instance_001.png",
"ref_video": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/structureEval/test/000015_Bravo_instance_001_camera.mp4",
"action_path": "/mnt/weka/home/hao.zhang/kaiqin/solaris/datasets/eval/structureEval/test/000015_Bravo_instance_001_camera.npy",
"num_inference_steps": 4,
"height": 352,
"width": 640,
"num_frames": 81
}
]
}