Document benchmark-contained evaluation asset layout
Browse files- benchmark/README.md +3 -3
benchmark/README.md
CHANGED
|
@@ -44,8 +44,8 @@ The tracking benchmark evaluates these fixed 512 sparse queries. The
|
|
| 44 |
intermediate 4096-query WorldTrack conversion files are not required for
|
| 45 |
evaluation and are intentionally omitted.
|
| 46 |
|
| 47 |
-
Model inputs and the remaining task ground truth are included under
|
| 48 |
-
`challenge_eval`
|
| 49 |
|
| 50 |
- `<variant>/<scene>/mp4/<sequence>.mp4`: 512 model-input videos;
|
| 51 |
- `<variant>/<scene>/exr_layers/depth/<sequence>/`: the 32 evaluated depth EXRs;
|
|
@@ -58,7 +58,7 @@ git clone https://github.com/jzr99/Syn4D_benchmark.git
|
|
| 58 |
cd Syn4D_benchmark
|
| 59 |
python -m pip install -r requirements.txt
|
| 60 |
python download_benchmark.py --output data/release
|
| 61 |
-
export SYN4D_DATA_ROOT="$PWD/data/release/challenge_eval"
|
| 62 |
export SYN4D_TRACKING_GT="$PWD/data/release/benchmark/data/tracking_gt"
|
| 63 |
```
|
| 64 |
|
|
|
|
| 44 |
intermediate 4096-query WorldTrack conversion files are not required for
|
| 45 |
evaluation and are intentionally omitted.
|
| 46 |
|
| 47 |
+
Model inputs and the remaining task ground truth are included under this
|
| 48 |
+
directory's `challenge_eval` subtree:
|
| 49 |
|
| 50 |
- `<variant>/<scene>/mp4/<sequence>.mp4`: 512 model-input videos;
|
| 51 |
- `<variant>/<scene>/exr_layers/depth/<sequence>/`: the 32 evaluated depth EXRs;
|
|
|
|
| 58 |
cd Syn4D_benchmark
|
| 59 |
python -m pip install -r requirements.txt
|
| 60 |
python download_benchmark.py --output data/release
|
| 61 |
+
export SYN4D_DATA_ROOT="$PWD/data/release/benchmark/challenge_eval"
|
| 62 |
export SYN4D_TRACKING_GT="$PWD/data/release/benchmark/data/tracking_gt"
|
| 63 |
```
|
| 64 |
|