TechJam2026-Jamlai-Bench/TechJam2026_Jamlai_Bench

TechJam2026 Jamlai Bench

This repository collects image data used to build the TechJam2026 Jamlai benchmark.

NTIRE / Fake

The ntire_fake configuration contains 50,000 NTIRE images labeled as fake. They are stored under NTIRE/Fake/ as self-contained Parquet shards with the original JPEG bytes embedded in the image column. Every selected image satisfies both width > 512 and height > 512.

After applying the strict dimension filter, the deterministic selection preserves each source labels.csv order:

• shard_5: 13,024 images
• shard_1: 23,732 images
• shard_2: 13,244 images

from datasets import load_dataset

dataset = load_dataset(
    "Edward0707/TechJam2026_Jamlai_Bench",
    "ntire_fake",
)

The per-sample provenance and SHA-256 values are recorded in NTIRE/Fake/selection_manifest.csv. Build and verification details are in NTIRE/Fake/summary.json.

Before redistributing these images, verify the source dataset's license and attribution requirements.