The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
EZPC - Pre-computed CLIP / SigLIP Embeddings
This dataset hosts the pre-computed image embeddings, cached text (classname / concept) embeddings, and class splits used in Explaining CLIP Zero-shot Predictions Through Concepts (CVPR 2026).
Running EZPC from scratch requires extracting CLIP / SigLIP features for every image in five benchmark datasets (CIFAR-100, CUB-200-2011, Places365, ImageNet, ImageNet-100) with multiple backbones. To skip the expensive feature-extraction step, we release the exact tensors we used in the paper - drop them into the EZPC repo's data/ folder and you can train, evaluate, and reproduce quantitative results without touching the raw images.
- π Paper: arXiv:2603.28211
- π» Code: github.com/oonat/ezpc
- π Project page: oonat.github.io/ezpc
- π€ Checkpoints: oonat/ezpc-checkpoints
What's Included
For each of the five datasets and each supported backbone, this repository provides:
- Full train / test embeddings + targets: CLIP/SigLIP image features for every training and test image across all classes, along with their class indices. Useful as a general-purpose feature dump and as the input to
data/split_dataset.pyif you want to regenerate the splits under a different seed or ratio. - Seen-split train / test embeddings: embeddings restricted to the 80% "seen" classes. The train split is used to fit the EZPC projection matrix A; the test split measures GZSL seen-class accuracy.
- Unseen-split train / test embeddings: embeddings restricted to the 20% "unseen" classes held out from training. Used for GZSL unseen-class evaluation (and for the
--target_datasetside of cross-dataset transfer). - Split-aligned target files:
seen_train_ids.pt,seen_test_ids.pt,unseen_train_ids.pt,unseen_test_ids.pt, each giving the class index for every row of the corresponding split. - Class split file: the exact seen / unseen class partition (seed 42, 80/20) used in the paper.
- Cached text embeddings: L2-normalized
"a photo of {x}"text embeddings for class names ({backbone}_classname_embs.pt, shape(C, d)) and concepts ({backbone}_concept_matrix.pt, shape(m, d)).test.pyloads these automatically so evaluation reproduces the reported numbers exactly and independent of GPU / CUDA version β without re-running the CLIP/SigLIP text encoder.
Supported backbones: CLIP RN50, CLIP ViT-B/32, CLIP ViT-L/14, SigLIP ViT-SO400M/14.
Repository Layout
Each dataset follows the same two-level structure: top-level embeddings/ holds the full image feature dump (all classes) plus the cached text embeddings, and embeddings/splits/ holds the seen/unseen partition used throughout the paper.
data/
βββ CIFAR-100/
β βββ embeddings/
β βββ RN50_train_embeddings.pt # (N_train, d) - all classes
β βββ RN50_test_embeddings.pt # (N_test, d) - all classes
β βββ ViT-B-32_train_embeddings.pt
β βββ ViT-B-32_test_embeddings.pt
β βββ ViT-L-14_train_embeddings.pt
β βββ ViT-L-14_test_embeddings.pt
β βββ siglip-so400m-patch14-384_train_embeddings.pt
β βββ siglip-so400m-patch14-384_test_embeddings.pt
β βββ RN50_classname_embs.pt # (C, d) - class-name text embeddings
β βββ RN50_concept_matrix.pt # (m, d) - concept text embeddings
β βββ ViT-B-32_classname_embs.pt
β βββ ViT-B-32_concept_matrix.pt
β βββ ViT-L-14_classname_embs.pt
β βββ ViT-L-14_concept_matrix.pt
β βββ siglip-so400m-patch14-384_classname_embs.pt
β βββ siglip-so400m-patch14-384_concept_matrix.pt
β βββ train_ids.pt # (N_train,) - class indices
β βββ test_ids.pt # (N_test,) - class indices
β βββ splits/
β βββ class_split.pt # {seen_classes, unseen_classes}
β βββ RN50_seen_train_embs.pt
β βββ RN50_seen_test_embs.pt
β βββ RN50_unseen_train_embs.pt
β βββ RN50_unseen_test_embs.pt
β βββ ViT-B-32_seen_train_embs.pt
β βββ ViT-B-32_seen_test_embs.pt
β βββ ViT-B-32_unseen_train_embs.pt
β βββ ViT-B-32_unseen_test_embs.pt
β βββ ViT-L-14_seen_train_embs.pt
β βββ ViT-L-14_seen_test_embs.pt
β βββ ViT-L-14_unseen_train_embs.pt
β βββ ViT-L-14_unseen_test_embs.pt
β βββ siglip-so400m-patch14-384_seen_train_embs.pt
β βββ siglip-so400m-patch14-384_seen_test_embs.pt
β βββ siglip-so400m-patch14-384_unseen_train_embs.pt
β βββ siglip-so400m-patch14-384_unseen_test_embs.pt
β βββ seen_train_ids.pt
β βββ seen_test_ids.pt
β βββ unseen_train_ids.pt
β βββ unseen_test_ids.pt
βββ CUB-200-2011/
β βββ embeddings/ ...
βββ ImageNet/
β βββ embeddings/ ...
βββ ImageNet-100/
β βββ embeddings/ ...
βββ Places365/
βββ embeddings/ ...
Quickstart
1. Download the data
pip install huggingface-hub
# Grab everything
hf download oonat/ezpc-embeddings \
--repo-type dataset \
--local-dir data
# Or just one dataset
hf download oonat/ezpc-embeddings \
--repo-type dataset \
--local-dir data \
--include "CIFAR-100/*"
# Or only what EZPC needs to evaluate: the GZSL splits + cached text embeddings
hf download oonat/ezpc-embeddings \
--repo-type dataset \
--local-dir data \
--include "*/embeddings/splits/*" "*/embeddings/*_classname_embs.pt" "*/embeddings/*_concept_matrix.pt"
2. Plug it into EZPC
Clone the EZPC GitHub repo and point --dataset_root at the folder you downloaded:
git clone https://github.com/oonat/ezpc.git
cd ezpc
conda env create -f environment.yml
conda activate ezpc
pip install -e .
# data/ lives next to train.py now - run training:
python train.py \
--dataset CIFAR-100 \
--dataset_root ./data \
--backbone RN50 \
--lambda_weight 1.0 \
--lr 0.01 \
--num_epochs 10000
Citation
If you use this dataset, please cite:
@InProceedings{Ozdemir_2026_CVPR,
author = {Ozdemir, Onat and Christensen, Anders and Alaniz, Stephan and Akata, Zeynep and Akbas, Emre},
title = {Explaining CLIP Zero-shot Predictions Through Concepts},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2026},
pages = {31336-31345}
}
Please also cite the original datasets whose images were used to produce these embeddings (CIFAR-100, CUB-200-2011, Places365, ImageNet, ImageNet-100) and the backbone models (CLIP, SigLIP).
Acknowledgements
- Concept vocabularies and class label mappings are taken from Label-free Concept Bottleneck Models. We thank the authors for open-sourcing them.
- Backbones from OpenAI CLIP, OpenCLIP, and Google SigLIP.
License
Released under the MIT License.
Note that these embeddings are derived from CIFAR-100, CUB-200-2011, Places365, ImageNet, and ImageNet-100. Users are responsible for complying with the original license and terms of use of those datasets, which may restrict commercial use β notably ImageNet and CUB-200-2011, which are released for non-commercial research only.
- Downloads last month
- 116