Pre-computed High-Dimensional Forensic Latents for ultra-fast Diffusion-Diff forensics training. - imgid (string): Image identifier. - label (int64): Class label (0: authentic, 1: fully synthetic, 2: tampered/inpainted).
Rows126,592
Configurations1
Size18.3 GB
Licenseapache-2.0
AccessPublicly accessible
Monthly Downloads144
Dataset Card
By Truong, published under apache-2.0, revision 2a209daab74d.
Pre-computed High-Dimensional Forensic Latents for ultra-fast Diffusion-Diff forensics training.
Dataset Details
- Source Dataset:
KhangTruong/COCO-inpainted - Model Extractor:
diffusion_diff_minimized - Cached Channels:
84 - Latent Resolution:
64x64(input512x512) - Format: Parquet with zstd compression
- Status:
in_progress
Schema / Columns
img_id(string): Image identifier.label(int64): Class label (0: authentic,1: fully synthetic,2: tampered/inpainted).z_high_dim(binary): Pre-computed 84-channel fp16 latent tensor bytes.mask(binary): Ground-truth binary manipulation mask encoded as PNG bytes.channels(int64): Total channel count (84).latent_h(int64): Latent height (64).latent_w(int64): Latent width (64).
High-Speed Training with SID-UNet
sid-train --config configs/experiments/diffusion_diff_minimized/default.yaml \
--cached-hf-repo KhangTruong/COCO-inpainted-cache
Structure
default 126,592 rows
| Split | Rows | Size |
|---|---|---|
| train | 122,184 | 84.3 GB |
| validation | 4,408 | 3.0 GB |
img_idstringlabelint64z_high_dimbinarymaskbinarychannelsint64latent_hint64latent_wint64
Details
- Repository
- KhangTruong/COCO-inpainted-cache
- Publisher
- Truong
- Task category
- Image segmentation
- Tags
- synthetic-image-detection, diffusion-diff, forensic-latents
- Size category
- Not stated by the source
- Languages
- Not stated by the source
- Revision
- 2a209daab74d13d0436a2493c190e7d6328d1350
- Last updated
- 2026-10-13
Files
63 files, 18.3 GB in total.
Data61 files · 18.3 GB
Documentation1 file · 1.3 KB
Repository1 file · 2.5 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| cache_info.json | Data | 561 B | — |
| train/train-00000.parquet | Data | 304.1 MB | 2a75e77d8647 |
| train/train-00001.parquet | Data | 304.2 MB | 31c8841abfce |
| train/train-00002.parquet | Data | 304.1 MB | d8fc5648807b |
| train/train-00003.parquet | Data | 304.8 MB | 8b36abe03465 |
| train/train-00004.parquet | Data | 304.8 MB | 3a75382b28b3 |
| train/train-00005.parquet | Data | 304.8 MB | 2a4a74ba8b33 |
| train/train-00006.parquet | Data | 304.8 MB | d88ec7f2c998 |
| train/train-00007.parquet | Data | 304.7 MB | 7622dbfbb9d9 |
| train/train-00008.parquet | Data | 304.7 MB | 9f2b00ffc703 |
| train/train-00009.parquet | Data | 304.7 MB | dd2d06942bb8 |
| train/train-00010.parquet | Data | 304.6 MB | e5f9c41c8b1d |
| train/train-00011.parquet | Data | 304.7 MB | 5cd42cd961c2 |
| train/train-00012.parquet | Data | 304.8 MB | 209021aa0373 |
| train/train-00013.parquet | Data | 304.7 MB | ab6ecfcd9272 |
| train/train-00014.parquet | Data | 304.7 MB | 43303b74f485 |
| train/train-00015.parquet | Data | 304.7 MB | 419aba84181d |
| train/train-00016.parquet | Data | 304.8 MB | a20aa476a06b |
| train/train-00017.parquet | Data | 304.7 MB | 509980047efd |
| train/train-00018.parquet | Data | 304.7 MB | 32462f61a388 |
| train/train-00019.parquet | Data | 304.7 MB | 1dbfd09e1927 |
| train/train-00020.parquet | Data | 304.8 MB | 604e5266436b |
| train/train-00021.parquet | Data | 304.8 MB | 590ffad9d4b8 |
| train/train-00022.parquet | Data | 306.2 MB | 3cfd1316b603 |
| train/train-00023.parquet | Data | 306.1 MB | af407b1c9e3b |
| train/train-00024.parquet | Data | 306.1 MB | f36f09538cb3 |
| train/train-00025.parquet | Data | 306.1 MB | 55ef9e64ddaa |
| train/train-00026.parquet | Data | 306.1 MB | c9d2c528767a |
| train/train-00027.parquet | Data | 306.1 MB | d5d94a570bfd |
| train/train-00028.parquet | Data | 306.1 MB | 7250a0f079d6 |
| train/train-00029.parquet | Data | 306.1 MB | 3a3953da725b |
| train/train-00030.parquet | Data | 306.0 MB | 801bcb11d079 |
| train/train-00031.parquet | Data | 306.1 MB | a1154d87d051 |
| train/train-00032.parquet | Data | 306.2 MB | d3e62e9f7420 |
| train/train-00033.parquet | Data | 306.1 MB | 031472b9123a |
| train/train-00034.parquet | Data | 306.0 MB | 5ffb68b6b2cc |
| train/train-00035.parquet | Data | 306.1 MB | 71cbe3953eee |
| train/train-00036.parquet | Data | 306.1 MB | d4d095074312 |
| train/train-00037.parquet | Data | 306.1 MB | 5e0b559fc2c0 |
| train/train-00038.parquet | Data | 306.2 MB | c9b22e6de66b |
| train/train-00039.parquet | Data | 306.1 MB | 0966b13ba9e8 |
| train/train-00040.parquet | Data | 306.1 MB | 6697c4de2873 |
| train/train-00041.parquet | Data | 306.1 MB | 2292e01a8200 |
| train/train-00042.parquet | Data | 306.1 MB | 29e0efef8605 |
| train/train-00043.parquet | Data | 306.1 MB | efc493ff139f |
| train/train-00044.parquet | Data | 306.1 MB | 4e9ee9d618c4 |
| train/train-00045.parquet | Data | 306.0 MB | 81d5be78a4bf |
| train/train-00046.parquet | Data | 306.0 MB | 01031ba17a65 |
| train/train-00047.parquet | Data | 306.2 MB | 7bbc89f45f4c |
| train/train-00048.parquet | Data | 306.2 MB | 6e34d1f5e4f4 |
| train/train-00049.parquet | Data | 306.1 MB | d6bfb79e6604 |
| train/train-00050.parquet | Data | 306.2 MB | 4073ad60008d |
| train/train-00051.parquet | Data | 306.0 MB | d77837351459 |
| train/train-00052.parquet | Data | 306.2 MB | 2175b60cb60f |
| train/train-00053.parquet | Data | 306.1 MB | a9a62cf17111 |
| train/train-00054.parquet | Data | 306.1 MB | 25e98a291df7 |
| train/train-00055.parquet | Data | 306.1 MB | 3c1df0d4299b |
| train/train-00056.parquet | Data | 306.1 MB | e4fa8ff3f140 |
| train/train-00057.parquet | Data | 306.1 MB | b01043f4d8c4 |
| train/train-00058.parquet | Data | 306.1 MB | bf565887320b |
| train/train-00059.parquet | Data | 306.2 MB | 531c36ad7cba |
| README.md | Documentation | 1.3 KB | — |
| .gitattributes | Repository | 2.5 KB | — |
License and Download
- License
- apache-2.0
- Access
- No access gate
Download from Truong
Released by Truong through its official repository on Hugging Face. Read the license.