Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,34 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: mit
|
| 3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
---
|
| 4 |
+
|
| 5 |
+
## CyberOrigin Dataset
|
| 6 |
+
|
| 7 |
+
Our data includes information from home services, the logistics industry, and laboratory scenarios.
|
| 8 |
+
For more details, please refer to our [Offical Data Website](https://cyberorigin2077.github.io/)
|
| 9 |
+
|
| 10 |
+
contents of the dataset:
|
| 11 |
+
```
|
| 12 |
+
cyber_take_the_item # dataset root path
|
| 13 |
+
└── data/
|
| 14 |
+
├── metadata_ID1_240808.json
|
| 15 |
+
├── segment_ids_ID1_240808.bin # for each frame segment_ids uniquely points to the segment index that frame i came from. You may want to use this to separate non-contiguous frames from different videos (transitions).
|
| 16 |
+
├── videos_ID1_240808.bin # 16x16 image patches at 30hz, each patch is vector-quantized into 2^18 possible integer values. These can be decoded into 256x256 RGB images using the provided magvit2.ckpt weights.
|
| 17 |
+
├── ...
|
| 18 |
+
└── ...
|
| 19 |
+
```
|
| 20 |
+
|
| 21 |
+
```json
|
| 22 |
+
{
|
| 23 |
+
"task": "Take the Item",
|
| 24 |
+
"total_episodes": 18779,
|
| 25 |
+
"total_frames": 3617759,
|
| 26 |
+
"token_dtype": "uint32",
|
| 27 |
+
"vocab_size": 262144,
|
| 28 |
+
"fps": 30,
|
| 29 |
+
"manipulation_type": "Single Arm",
|
| 30 |
+
"language_annotation": "None",
|
| 31 |
+
"scene_type": "Table Top",
|
| 32 |
+
"data_collect_method": "Directly Collection on Human"
|
| 33 |
+
}
|
| 34 |
+
```
|