Rename to OmniPro, update task abbreviations, add eval code link
Browse files
README.md
CHANGED
|
@@ -1,6 +1,6 @@
|
|
| 1 |
-
#
|
| 2 |
|
| 3 |
-
This is a **lightweight subset** of [
|
| 4 |
|
| 5 |
## Contents
|
| 6 |
|
|
@@ -8,6 +8,20 @@ This is a **lightweight subset** of [OmniProact-Bench](https://huggingface.co/da
|
|
| 8 |
- **245 source videos** corresponding to the sampled annotations
|
| 9 |
- `benchmark_subset.json`: all 270 evaluation samples with full annotations
|
| 10 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
## Sampling Method
|
| 12 |
|
| 13 |
Random sampling with seed=42, stratified by task type (30 samples per task), preserving the original distribution of audio dependencies and trigger types.
|
|
@@ -16,7 +30,9 @@ Random sampling with seed=42, stratified by task type (30 samples per task), pre
|
|
| 16 |
|
| 17 |
This subset is intended solely for quick quality inspection. The complete benchmark (2,700 samples, 1,262 videos, 32.6 GB) is available at:
|
| 18 |
|
| 19 |
-
👉 [omniproact-bench/
|
|
|
|
|
|
|
| 20 |
|
| 21 |
## File Structure
|
| 22 |
|
|
|
|
| 1 |
+
# OmniPro Subset (Review Sample)
|
| 2 |
|
| 3 |
+
This is a **lightweight subset** of [OmniPro](https://huggingface.co/datasets/omniproact-bench/OmniPro) for convenient reviewer inspection.
|
| 4 |
|
| 5 |
## Contents
|
| 6 |
|
|
|
|
| 8 |
- **245 source videos** corresponding to the sampled annotations
|
| 9 |
- `benchmark_subset.json`: all 270 evaluation samples with full annotations
|
| 10 |
|
| 11 |
+
## Task Types
|
| 12 |
+
|
| 13 |
+
| Task | Abbr. | Samples |
|
| 14 |
+
| --------------------------- | ------------- | ------- |
|
| 15 |
+
| Instant Event Alert | Event-Alert | 30 |
|
| 16 |
+
| Semantic Condition Alert | Cond.-Alert | 30 |
|
| 17 |
+
| Explicit Target Grounding | Target-Ground | 30 |
|
| 18 |
+
| Snapshot Counting | Snap.-Count | 30 |
|
| 19 |
+
| Cumulative Counting | Cum.-Count | 30 |
|
| 20 |
+
| Dedup Counting | Dedup.-Count | 30 |
|
| 21 |
+
| Realtime State Monitor | State-Monitor | 30 |
|
| 22 |
+
| Event Narration | Event-Narr. | 30 |
|
| 23 |
+
| Sequential Step Instruction | Step-Inst. | 30 |
|
| 24 |
+
|
| 25 |
## Sampling Method
|
| 26 |
|
| 27 |
Random sampling with seed=42, stratified by task type (30 samples per task), preserving the original distribution of audio dependencies and trigger types.
|
|
|
|
| 30 |
|
| 31 |
This subset is intended solely for quick quality inspection. The complete benchmark (2,700 samples, 1,262 videos, 32.6 GB) is available at:
|
| 32 |
|
| 33 |
+
👉 [omniproact-bench/OmniPro](https://huggingface.co/datasets/omniproact-bench/OmniPro)
|
| 34 |
+
|
| 35 |
+
Evaluation code is available at: **[🔗 OmniPro Evaluation Code](https://anonymous.4open.science/r/OmniPro-5031)**
|
| 36 |
|
| 37 |
## File Structure
|
| 38 |
|