diff --git a/src/gasbench/dataset/configs/real_audio.yaml b/src/gasbench/dataset/configs/real_audio.yaml index f36703d..a29ac13 100644 --- a/src/gasbench/dataset/configs/real_audio.yaml +++ b/src/gasbench/dataset/configs/real_audio.yaml @@ -755,3 +755,57 @@ datasets: generator_family: real content_category: speech notes: CML-TTS Spanish train split, released from v22 bmcore holdouts + + # ── Released v23 holdouts — deferred to upstream repos (no 34data mirror) ── + + - name: cml-tts-dutch + path: ylacombe/cml-tts + hf_revision: 48aedd7a8ea08efc75cae3e81fbefa9720df0812 + modality: audio + media_type: real + source_format: parquet + media_per_archive: 500 + archives_per_dataset: 2 + hf_subfolders: ['dutch'] + include_paths: ['train-'] + generator_family: real + content_category: speech + + - name: cml-tts-italian + path: ylacombe/cml-tts + hf_revision: 48aedd7a8ea08efc75cae3e81fbefa9720df0812 + modality: audio + media_type: real + source_format: parquet + media_per_archive: 500 + archives_per_dataset: 2 + hf_subfolders: ['italian'] + include_paths: ['train-'] + generator_family: real + content_category: speech + + - name: cml-tts-polish + path: ylacombe/cml-tts + hf_revision: 48aedd7a8ea08efc75cae3e81fbefa9720df0812 + modality: audio + media_type: real + source_format: parquet + media_per_archive: 500 + archives_per_dataset: 2 + hf_subfolders: ['polish'] + include_paths: ['train-'] + generator_family: real + content_category: speech + + - name: cml-tts-portuguese + path: ylacombe/cml-tts + hf_revision: 48aedd7a8ea08efc75cae3e81fbefa9720df0812 + modality: audio + media_type: real + source_format: parquet + media_per_archive: 500 + archives_per_dataset: 2 + hf_subfolders: ['portuguese'] + include_paths: ['train-'] + generator_family: real + content_category: speech diff --git a/src/gasbench/dataset/configs/real_images.yaml b/src/gasbench/dataset/configs/real_images.yaml index 39f68f3..c9b5452 100644 --- a/src/gasbench/dataset/configs/real_images.yaml +++ b/src/gasbench/dataset/configs/real_images.yaml @@ -1223,3 +1223,29 @@ datasets: archives_per_dataset: 1 generator_family: real content_category: medical + + # ── Released v23 holdouts — deferred to upstream repos (no 34data mirror) ── + + - name: conceptual-captions-0-1m + path: obscure-entropy/conceptual_captions_hu_filtered + hf_revision: e70871e77287ca4aec4a12ff2f9cec32cad5694a + modality: image + media_type: real + source_format: parquet + media_per_archive: 1000 + archives_per_dataset: 1 + include_paths: ['data/conceptual_captions_0_1m.parquet'] + generator_family: real + content_category: diverse + + - name: rocov2-radiology + path: eltorio/ROCOv2-radiology + hf_revision: 80ffeef4eb8d34d27cb5c2815305f1d8aee8a83c + modality: image + media_type: real + source_format: parquet + media_per_archive: 1000 + archives_per_dataset: 1 + include_paths: ['data/train-'] + generator_family: real + content_category: medical diff --git a/src/gasbench/dataset/configs/real_videos.yaml b/src/gasbench/dataset/configs/real_videos.yaml index a0fc879..f006803 100644 --- a/src/gasbench/dataset/configs/real_videos.yaml +++ b/src/gasbench/dataset/configs/real_videos.yaml @@ -1093,3 +1093,65 @@ datasets: generator_family: real content_category: action notes: Released from v22 bmcore holdouts + + # ── Released v23 holdouts — deferred to upstream repos (no 34data mirror) ── + + - name: timelens2-93k-00003 + path: MCG-NJU/TimeLens2-93K + hf_revision: fce4049cdf5d720e2651b956c58e664c3386c021 + modality: video + media_type: real + source_format: tar + media_per_archive: 500 + archives_per_dataset: 1 + include_paths: ['videos/videos-00003-of-00018.tar'] + generator_family: real + content_category: diverse + + - name: llava-onevision-2-mid-training-00000 + path: mvp-lab/LLaVA-OneVision-2-Data + hf_revision: bf7038ba5d17771dc113a3072f98591ad2e29541 + modality: video + media_type: real + source_format: tar + media_per_archive: 500 + archives_per_dataset: 1 + include_paths: ['mid_training_video/30s/train_00000_of_09707.tar'] + generator_family: real + content_category: diverse + + - name: rekadaily-phone-shard-00000 + path: RekaAI/RekaDaily-10k-raw + hf_revision: a42e2da9aaeef7c8653d9de50772c97ceb954251 + modality: video + media_type: real + source_format: tar + media_per_archive: 500 + archives_per_dataset: 1 + include_paths: ['data/video_capture_first_person_videos_phone/shard-00000.tar'] + generator_family: real + content_category: action + + - name: rekadaily-phone-shard-00001 + path: RekaAI/RekaDaily-10k-raw + hf_revision: a42e2da9aaeef7c8653d9de50772c97ceb954251 + modality: video + media_type: real + source_format: tar + media_per_archive: 500 + archives_per_dataset: 1 + include_paths: ['data/video_capture_first_person_videos_phone/shard-00001.tar'] + generator_family: real + content_category: action + + - name: rekadaily-phone-shard-00002 + path: RekaAI/RekaDaily-10k-raw + hf_revision: a42e2da9aaeef7c8653d9de50772c97ceb954251 + modality: video + media_type: real + source_format: tar + media_per_archive: 500 + archives_per_dataset: 1 + include_paths: ['data/video_capture_first_person_videos_phone/shard-00002.tar'] + generator_family: real + content_category: action diff --git a/src/gasbench/dataset/configs/synthetic_images.yaml b/src/gasbench/dataset/configs/synthetic_images.yaml index ef12539..76da567 100644 --- a/src/gasbench/dataset/configs/synthetic_images.yaml +++ b/src/gasbench/dataset/configs/synthetic_images.yaml @@ -974,3 +974,128 @@ datasets: archives_per_dataset: 1 generator_family: mixed-generation content_category: diverse + + # ── Released v23 holdouts — deferred to upstream repos (no 34data mirror) ── + + - name: swappir-gpen-celeba-hq + path: Sumsub/Swappir + hf_revision: 02bf824fdd87577979105718d066d30aae780dca + modality: image + media_type: synthetic + source_format: zip + media_per_archive: 1000 + archives_per_dataset: 1 + include_paths: ['GPEN_CelebA_HQ.zip'] + generator_family: gpen + content_category: faces + + - name: swappir-gpen-fairface + path: Sumsub/Swappir + hf_revision: 02bf824fdd87577979105718d066d30aae780dca + modality: image + media_type: synthetic + source_format: zip + media_per_archive: 1000 + archives_per_dataset: 1 + include_paths: ['GPEN_fairface.zip'] + generator_family: gpen + content_category: faces + + - name: swappir-celeba-hq-roop + path: Sumsub/Swappir + hf_revision: 02bf824fdd87577979105718d066d30aae780dca + modality: image + media_type: semisynthetic + source_format: zip + media_per_archive: 1000 + archives_per_dataset: 1 + include_paths: ['CelebA_HQ_roop.zip'] + exclude_paths: ['GPEN_'] + generator_family: faceswap + content_category: faces + + - name: swappir-celeba-hq-simswap + path: Sumsub/Swappir + hf_revision: 02bf824fdd87577979105718d066d30aae780dca + modality: image + media_type: semisynthetic + source_format: zip + media_per_archive: 1000 + archives_per_dataset: 1 + include_paths: ['CelebA_HQ_simswap.zip'] + exclude_paths: ['GPEN_'] + generator_family: faceswap + content_category: faces + + - name: swappir-fairface-roop + path: Sumsub/Swappir + hf_revision: 02bf824fdd87577979105718d066d30aae780dca + modality: image + media_type: semisynthetic + source_format: zip + media_per_archive: 1000 + archives_per_dataset: 1 + include_paths: ['fairface_roop.zip'] + exclude_paths: ['GPEN_'] + generator_family: faceswap + content_category: faces + + - name: scaledf-ffpp-deepfakes + path: scaledf/ScaleDF + hf_revision: fdc0f92f1f133d5c8742578917aa280b32d9d8ce + modality: image + media_type: semisynthetic + source_format: tar + media_per_archive: 1000 + archives_per_dataset: 1 + include_paths: ['ScaleDF/train/FFpp_Deepfakes.tar'] + generator_family: faceswap + content_category: faces + + - name: scaledf-ffpp-face2face + path: scaledf/ScaleDF + hf_revision: fdc0f92f1f133d5c8742578917aa280b32d9d8ce + modality: image + media_type: semisynthetic + source_format: tar + media_per_archive: 1000 + archives_per_dataset: 1 + include_paths: ['ScaleDF/train/FFpp_Face2Face.tar'] + generator_family: face2face + content_category: faces + + - name: scaledf-e4e + path: scaledf/ScaleDF + hf_revision: fdc0f92f1f133d5c8742578917aa280b32d9d8ce + modality: image + media_type: semisynthetic + source_format: tar + media_per_archive: 1000 + archives_per_dataset: 1 + include_paths: ['ScaleDF/train/E4E_faces.tar'] + generator_family: e4e + content_category: faces + + - name: scaledf-diffusionclip + path: scaledf/ScaleDF + hf_revision: fdc0f92f1f133d5c8742578917aa280b32d9d8ce + modality: image + media_type: semisynthetic + source_format: tar + media_per_archive: 1000 + archives_per_dataset: 1 + include_paths: ['ScaleDF/train/DiffusionCLIP_faces.tar'] + generator_family: diffusionclip + content_category: faces + + - name: dgm4-styleclip + path: rshaojimmy/DGM4 + hf_revision: 9efd3f1ca14cb1dac1692cdce0beea6844b968a6 + modality: image + media_type: semisynthetic + source_format: zip + media_per_archive: 1000 + archives_per_dataset: 1 + include_paths: ['manipulation/StyleCLIP.zip'] + generator_family: styleclip + content_category: faces diff --git a/src/gasbench/dataset/configs/synthetic_videos.yaml b/src/gasbench/dataset/configs/synthetic_videos.yaml index eb997b8..4758c89 100644 --- a/src/gasbench/dataset/configs/synthetic_videos.yaml +++ b/src/gasbench/dataset/configs/synthetic_videos.yaml @@ -1338,3 +1338,17 @@ datasets: archives_per_dataset: 1 generator_family: float content_category: faces + + # ── Released v23 holdouts — deferred to upstream repos (no 34data mirror) ── + + - name: tip-i2v-i2vgenxl + path: WenhaoWang/TIP-I2V + hf_revision: 958f51df0ba725f53e2387bc19011a001d5ce7a4 + modality: video + media_type: synthetic + source_format: tar + media_per_archive: 500 + archives_per_dataset: 1 + include_paths: ['subset_videos_tar/i2vgenxl_videos_subset_1.tar'] + generator_family: i2vgenxl + content_category: diverse