Dataset Viewer
The dataset viewer is not available for this split.
The size of the content of the first rows (352246 B) exceeds the maximum supported size (200000 B) even after truncation. Please report the issue.
Error code: TooBigContentError
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
AIC 2026 Multi-Modal Retrieval Artifacts (Full 873 Videos)
This repository contains the complete preprocessed production artifacts for the 873-video dataset (130 hours) for AIC 2026.
📦 Contents:
siglip2_sq8.index(108 MB): FAISS SQ8 quantized index for 98,647 vectors (google/siglip2-so400m-patch14-384, dim 1152).siglip2_embeddings.npy(433 MB): Raw float32 SigLIP2 image embeddings for all 98,647 keyframes.retained_keyframes.json(15 MB): Frame mapping for all 96,340 retained keyframes across 873 videos.asr_transcripts.json(8.4 MB): 14,681 transcribed speech segments (Whisper).ocr_detections.json(10.6 MB): 140,411 masked text detections (EasyOCR).yolo_detections.json(79.9 MB): 523,965 detected bounding boxes (YOLOv11).metadata.sqlite(147 MB): Unified relational SQLite database with full-text search (FTS5 tokenization).
- Downloads last month
- 118