koukyo1994
commited on
Commit
•
1ed4199
1
Parent(s):
950a943
update README.md
Browse files
README.md
CHANGED
@@ -38,6 +38,14 @@ tags:
|
|
38 |
Terra generates video continuations based on short video clips of approximately three frames and trajectory instructions.
|
39 |
A key feature of Terra is its **high adherence to trajectory instructions**, enabling accurate and reliable action-conditioned video generation.
|
40 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
41 |
## Related Links
|
42 |
|
43 |
For more technical details and discussions, please refer to:
|
|
|
38 |
Terra generates video continuations based on short video clips of approximately three frames and trajectory instructions.
|
39 |
A key feature of Terra is its **high adherence to trajectory instructions**, enabling accurate and reliable action-conditioned video generation.
|
40 |
|
41 |
+
We have developed two versions of the Terra model to date. The `v1` model, as detailed in the paper, exhibits a bias towards generating videos that veer to the right. To address this issue, we introduced the `v2` model, incorporating slight architectural modifications to mitigate this tendency and produce more balanced outputs. The performance of each model is outlined below.
|
42 |
+
|
43 |
+
||Vista|Terra(v1)|Terra(v2)|
|
44 |
+
|---|---|---|---|
|
45 |
+
|Accuracy (↑)| 0.307 |0.441 | **0.632** |
|
46 |
+
|ADE (↓)|4.50 | 3.98 | **3.86** |
|
47 |
+
|FDE (↓)|8.66|8.21| **8.05**|
|
48 |
+
|
49 |
## Related Links
|
50 |
|
51 |
For more technical details and discussions, please refer to:
|