duyaaa commited on
Commit
bff6ef3
·
verified ·
1 Parent(s): 161c599

Add dataset card and usage documentation

Browse files
Files changed (1) hide show
  1. README.md +88 -0
README.md ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ task_categories:
4
+ - video-to-text
5
+ - feature-extraction
6
+ - text-to-image
7
+ tags:
8
+ - video-retrieval
9
+ - aic2026
10
+ - siglip2
11
+ - faiss
12
+ - vietnamese
13
+ size_categories:
14
+ - 1GB<n<10GB
15
+ ---
16
+
17
+ # AIC-2026: 15-Video Pilot Dataset & Preprocessing Artifacts
18
+
19
+ Dataset phục vụ phát triển giao diện (UI) và tính năng truy vấn tìm kiếm (Search / Retrieval) cho cuộc thi AI Challenge 2026 (AIC-2026).
20
+
21
+ ## 📁 Cấu trúc Dataset
22
+
23
+ ```text
24
+ ├── videos/ # 15 video MP4 gốc (L21_V001 -> L21_V016)
25
+ │ ├── L21_V001.mp4
26
+ │ ├── ...
27
+ │ └── L21_V016.mp4
28
+ ├── indexes/ # Vector Index FAISS phục vụ visual search
29
+ │ ├── visual-sqfp16.faiss # Index Scalar Quantizer FP16 (41,730 vectors, 81.8MB)
30
+ │ ├── visual-sqfp16-manifest.json
31
+ │ ├── visual-flat.faiss # Index Flat L2/InnerProduct (163.3MB)
32
+ │ └── visual-flat-manifest.json
33
+ ├── canonical/ # Metadata Store & Identity
34
+ │ ├── metadata.sqlite # SQLite database lưu trữ keyframes, shots, transcripts
35
+ │ └── dataset-identity.json
36
+ ├── pilot-15v-canonical.tar.gz # Full archive chứa 42,000+ ảnh keyframes đã trích xuất
37
+ └── query_KIS.csv # 53 Benchmark test queries (Known-Item Search)
38
+ ```
39
+
40
+ ## 🚀 Hướng dẫn Sử dụng
41
+
42
+ ### 1. Tải Artifacts & Video về Local
43
+
44
+ ```python
45
+ from huggingface_hub import snapshot_download
46
+
47
+ snapshot_download(
48
+ repo_id="duyaaa/skibidi15",
49
+ repo_type="dataset",
50
+ local_dir="./data",
51
+ )
52
+ ```
53
+
54
+ ### 2. Cấu hình Môi trường cho Backend / API
55
+
56
+ Đặt các biến môi trường sau trong file `.env`:
57
+
58
+ ```bash
59
+ AIC_DATA_DIR="./data"
60
+ AIC_V2_ARTIFACTS_DIR="./data/canonical"
61
+ AIC_DB_PATH="./data/canonical/metadata.sqlite"
62
+ AIC_V2_INDEX_MODE="sqfp16"
63
+ AIC_V2_INDEX_PATH="./data/indexes/visual-sqfp16.faiss"
64
+ AIC_V2_INDEX_MANIFEST="./data/indexes/visual-sqfp16-manifest.json"
65
+ AIC_V2_ENABLE_SERVE="true"
66
+ ```
67
+
68
+ ### 3. Chạy Backend Search API
69
+
70
+ ```bash
71
+ uvicorn backend.main:app --host 0.0.0.0 --port 8000 --reload
72
+ ```
73
+
74
+ - API Endpoint Tìm kiếm: `POST http://localhost:8000/api/v2/search`
75
+ - Swagger Docs: `http://localhost:8000/docs`
76
+
77
+ ### 4. Đánh giá Benchmark 53 Query KIS
78
+
79
+ ```bash
80
+ python scripts/run_eval_direct.py
81
+ ```
82
+ - **Recall@1**: 56.6%
83
+ - **Recall@5**: 64.2%
84
+ - **Recall@20**: 84.9%
85
+ - **P50 Latency**: ~274 ms
86
+
87
+ ---
88
+ *Created for AIC-2026 Pipeline Development.*