andrewlababy commited on
Commit
b9eb9be
Β·
verified Β·
1 Parent(s): ac1a2f1

Update video specs for 1080p sample pack

Browse files
Files changed (1) hide show
  1. README.md +37 -24
README.md CHANGED
@@ -42,6 +42,8 @@ Key characteristics:
42
  - Intel RealSense D435I RGB-D capture for overhead view
43
  - Chinese cooking tasks involving tool use, bimanual coordination, and fine-grained food manipulation
44
 
 
 
45
  ---
46
 
47
  ## Sample Pack Contents
@@ -50,8 +52,8 @@ Key characteristics:
50
 
51
  | File | Description |
52
  |------|-------------|
53
- | `egocentric.mp4` | Head-mounted action camera, first-person view (3840Γ—2160, 29.97fps) |
54
- | `side_view.mp4` | Fixed side-view phone camera (1920Γ—1080, 60fps) |
55
  | `overhead/overhead.mp4` | Fixed overhead RealSense RGB (1280Γ—720, 15fps) |
56
  | `depth/depth.hdf5` | Aligned depth frames, float32 in meters (480Γ—848, 2379 frames) |
57
 
@@ -59,8 +61,8 @@ Key characteristics:
59
 
60
  | File | Description |
61
  |------|-------------|
62
- | `egocentric.mp4` | Head-mounted action camera, first-person view (3840Γ—2160, 29.97fps) |
63
- | `side_view.mp4` | Fixed side-view phone camera (1920Γ—1080, ~60fps) |
64
 
65
  > Depth sample is included for Task 1. Additional depth recordings may be available depending on the task and capture setup.
66
 
@@ -73,29 +75,28 @@ Key characteristics:
73
  ## File Structure
74
 
75
  ```
76
- samplepack_video/
77
- β”œβ”€β”€ task_01_cutting/
78
- β”‚ β”œβ”€β”€ egocentric.mp4
79
- β”‚ β”œβ”€β”€ side_view.mp4
80
- β”‚ β”œβ”€β”€ overhead/
81
- β”‚ β”‚ β”œβ”€β”€ overhead.mp4
82
- β”‚ β”‚ └── overhead_*.PNG
83
- β”‚ β”œβ”€β”€ depth/
84
- β”‚ β”‚ β”œβ”€β”€ depth.hdf5
85
- β”‚ β”‚ └── check_*.jpg
86
- β”‚ β”œβ”€β”€ egocentric_screenshot_*.PNG
87
- β”‚ └── side_view_*.PNG
88
- └── task_02_stir_fry/
89
- β”œβ”€β”€ egocentric.mp4
90
- β”œβ”€β”€ side_view.mp4
91
- β”œβ”€β”€ egocentric_screenshot_*.PNG
92
- └── side_view_*.PNG
93
  ```
94
 
95
  **Read depth data (Python):**
96
  ```python
97
  import h5py
98
- with h5py.File("samplepack_video/task_01_cutting/depth/depth.hdf5", "r") as f:
99
  depth = f["depth_meters"][:] # (2379, 480, 848) float32, meters
100
  ts = f["timestamps"][:]
101
  ```
@@ -104,10 +105,21 @@ with h5py.File("samplepack_video/task_01_cutting/depth/depth.hdf5", "r") as f:
104
 
105
  ## Camera Setup
106
 
 
 
 
 
 
 
 
 
 
 
 
107
  | View | Device | Resolution | Frame Rate |
108
  |------|--------|------------|------------|
109
- | Egocentric | Head-mounted action camera | 3840Γ—2160 (4K) | 29.97 fps |
110
- | Side | Fixed smartphone | 1920Γ—1080 | 60 fps |
111
  | Overhead RGB-D | Intel RealSense D435I | RGB 1280Γ—720 / Depth 848Γ—480 | 15 fps |
112
 
113
  ---
@@ -150,6 +162,7 @@ This public sample pack is intended for technical evaluation and early research
150
 
151
  Additional materials may be available upon request:
152
 
 
153
  - Longer multi-view MP4 recordings
154
  - Additional HDF5 metric depth sequences
155
  - RealSense raw `.bag` recordings
 
42
  - Intel RealSense D435I RGB-D capture for overhead view
43
  - Chinese cooking tasks involving tool use, bimanual coordination, and fine-grained food manipulation
44
 
45
+ > **Note:** Large egocentric videos in this sample pack are provided as **1080p preview encodings** for easier download. Full **4K** source recordings are available upon request.
46
+
47
  ---
48
 
49
  ## Sample Pack Contents
 
52
 
53
  | File | Description |
54
  |------|-------------|
55
+ | `egocentric_1080p.mp4` | Head-mounted action camera, first-person view (1920Γ—1080, 30fps preview) |
56
+ | `side_view.mp4` | Fixed side-view phone camera (1920Γ—1080, 60fps, original capture) |
57
  | `overhead/overhead.mp4` | Fixed overhead RealSense RGB (1280Γ—720, 15fps) |
58
  | `depth/depth.hdf5` | Aligned depth frames, float32 in meters (480Γ—848, 2379 frames) |
59
 
 
61
 
62
  | File | Description |
63
  |------|-------------|
64
+ | `egocentric_1080p.mp4` | Head-mounted action camera, first-person view (1920Γ—1080, 30fps preview) |
65
+ | `side_view.mp4` | Fixed side-view phone camera (1920Γ—1080, 30fps preview) |
66
 
67
  > Depth sample is included for Task 1. Additional depth recordings may be available depending on the task and capture setup.
68
 
 
75
  ## File Structure
76
 
77
  ```
78
+ task_01_cutting/
79
+ β”œβ”€β”€ egocentric_1080p.mp4
80
+ β”œβ”€β”€ side_view.mp4
81
+ β”œβ”€β”€ overhead/
82
+ β”‚ β”œβ”€β”€ overhead.mp4
83
+ β”‚ └── overhead_*.PNG
84
+ β”œβ”€β”€ depth/
85
+ β”‚ β”œβ”€β”€ depth.hdf5
86
+ β”‚ └── check_*.jpg
87
+ β”œβ”€β”€ egocentric_screenshot_*.PNG
88
+ └── side_view_*.PNG
89
+ task_02_stir_fry/
90
+ β”œβ”€β”€ egocentric_1080p.mp4
91
+ β”œβ”€β”€ side_view.mp4
92
+ β”œβ”€β”€ egocentric_screenshot_*.PNG
93
+ └── side_view_*.PNG
 
94
  ```
95
 
96
  **Read depth data (Python):**
97
  ```python
98
  import h5py
99
+ with h5py.File("task_01_cutting/depth/depth.hdf5", "r") as f:
100
  depth = f["depth_meters"][:] # (2379, 480, 848) float32, meters
101
  ts = f["timestamps"][:]
102
  ```
 
105
 
106
  ## Camera Setup
107
 
108
+ ### Sample pack (this repository)
109
+
110
+ | View | Task 1 | Task 2 |
111
+ |------|--------|--------|
112
+ | Egocentric | 1920Γ—1080, 30fps (`egocentric_1080p.mp4`) | 1920Γ—1080, 30fps (`egocentric_1080p.mp4`) |
113
+ | Side | 1920Γ—1080, 60fps (`side_view.mp4`) | 1920Γ—1080, 30fps (`side_view.mp4`) |
114
+ | Overhead RGB | 1280Γ—720, 15fps | β€” |
115
+ | Overhead depth | 848Γ—480 (HDF5) | β€” |
116
+
117
+ ### Original capture specs (full dataset on request)
118
+
119
  | View | Device | Resolution | Frame Rate |
120
  |------|--------|------------|------------|
121
+ | Egocentric | Head-mounted action camera (DJI Osmo Action 3) | 3840Γ—2160 (4K) | 29.97 fps |
122
+ | Side | Fixed smartphone | 1920Γ—1080 | up to 60 fps |
123
  | Overhead RGB-D | Intel RealSense D435I | RGB 1280Γ—720 / Depth 848Γ—480 | 15 fps |
124
 
125
  ---
 
162
 
163
  Additional materials may be available upon request:
164
 
165
+ - Full **4K** egocentric MP4 recordings
166
  - Longer multi-view MP4 recordings
167
  - Additional HDF5 metric depth sequences
168
  - RealSense raw `.bag` recordings