datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
features-dinov3-vith16plus-224-imagenet-22k-wdsCT-RATE_RAPTOR_DINOV3_Embedded_Validzangei-dit-stage-1-250k-256px-dinov3
Zangei 256px DINOv3 Features — Repacked
Repacked from the known Colab extraction layout.
Source dataset: kingsidharth/zangei-dit-stage-1-250k-256px-img
Feature model: DINOv3 ViT-S/16
Image variants:
resized
square
Feature kinds:
cls
reg
patch
Files:
shards/dinov3_vits16_256_.safetensors
Matching row index:
shards/dinov3_vits16_256_.parquet
Tensor keys:
cls
reg
patch
Manifests:
manifests/files.parquet
manifests/files.csv
config.json
cls_dinov3-vith16plus_in22kRetailGlue-lightglue-dinov3-vitlnyuv2-dinov3-featuresRetailGlue-lightglue-dinov3-vitbRetailGlue-lightglue-dinov3-vitskitti-dinov3-featureszangei-dit-stage-1-250k-128px-img-dinov3
Zangei 128px Images + DINOv3 Features
Prepared from the 256px Zangei stage-1 assets.
Files:
shards/images_uint8_128_*.safetensors
shards/dinov3_vits16_128_patch_*.safetensors
alignment/sample_ids.jsonl
manifests/files.parquet
manifests/asset_manifest.yaml
Image rows are aligned to text/training row order.
DINO rows are source dataset row order; use alignment/sample_ids.jsonl.
ricl_dinov3_embeds