datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
CLEVR-HOPE
CLEVR-HOPE
The CLEVR Held-Out Pair Evaluation (CLEVR-HOPE) dataset is a diagnostic dataset for testing the systematicity of VQA models.
CLEVR-HOPE is a controlled setting to test whether VQA models generalize to pairs of attribute values that were not seen during either training or fine-tuning.
Within CLEVR-HOPE, we refer to an unseen pair of attribute values as a Held-Out Pair (HOP). The dataset is composed of 29 sub-datasets, each for a different HOP.
For each of the 29 HOPs, we… See the full description on the dataset page: https://huggingface.co/datasets/user9000/CLEVR-HOPE.clevr-multichange
CLEVR-Multi-Change (30–40 objects)
Two-image change-captioning data used in "Stateful Visual Encoders for
Vision-Language Models" (the Multi-object Visual Differencing task). Each example is a before/after pair of a CLEVR scene
with 30–40 objects and 4 simultaneous changes (add / delete / move /
replace), rendered at 768×768 with a wide camera angle. Built with the
CLEVR-Multi-Change engine (Johnson et al. 2017; Qiu et al. 2021).
Code & paper:… See the full description on the dataset page: https://huggingface.co/datasets/zwcolin/clevr-multichange.Clevr_CoGenT_TrainA_70K_ComplexSuper-CLEVR
Super-CLEVR: A Virtual Benchmark to Diagnose Domain Robustness in Visual Reasoning
[CVPR 2023 Highlight (top 2.5%)]
Paper: Super-CLEVR: A Virtual Benchmark to Diagnose Domain Robustness in Visual Reasoning
Authors: Zhuowan Li, Xingrui Wang, Elias Stengel-Eskin, Adam Kortylewski, Wufei Ma, Benjamin Van Durme, Alan Yuille
Dataset Description
Super-CLEVR is a synthetic dataset designed to systematically study the domain robustness of visual reasoning models across… See the full description on the dataset page: https://huggingface.co/datasets/RyanWW/Super-CLEVR.clevr_cogen_a_trainClevr_CoGenT_ValAclevr-webdatasetclevr-mathclevrerwds_vtab-clevr_count_allmm_clevrclevr_count_70kThis dataset is borrowed from clevr_cogen_a_train
Clevr_Complex_70kCLEVR_30K_obj_2_3clevr-bboxCLEVR_categoriesClevr_CoGenT_TrainA_R1clevrCLEVR-BT-DB
How to install?
!pip install datasets -q
from huggingface_hub import snapshot_download
import pandas as pd
import matplotlib.pyplot as plt
# First step: download an entire datatset
snapshot_download(repo_id="Aborevsky01/CLEVR-BT-DB", repo_type="dataset", local_dir='path-to-your-local-dir')
# Second step: unarchive the images for VQA
!unzip [path-to-your-local-dir]/[type-of-task]/images.zip
# Example of the triplet (image - question -… See the full description on the dataset page: https://huggingface.co/datasets/Aborevsky01/CLEVR-BT-DB.wds_vtab-clevr_closest_object_distanceclevr_r1
CLEVR R1
CLEVR R1 is a multimodal reasoning dataset generated with R1 to distill multimodal reasoning abilities into models. This dataset was taken from MMInstruction/Clevr_CoGenT_TrainA_R1.
clevr-changeclevr_relspatial_clevr_numbered_2000Clevr_CoGenT_ValBclevr-cogent-r1sample_clevrclevr-tr
Dataset Card for CoT
Dataset Sources
Repository: LLaVA-CoT GitHub Repository
Paper: LLaVA-CoT on arXiv
Dataset Structure
Turkish - tr dataset
unzip image.zip
The train.jsonl file contains the question-answering data and is structured in the following format:
{
"id": "example_id",
"image": "example_image_path",
"conversations": [
{"from": "human", "value": "Lütfen resimdeki kırmızı metal nesnelerin sayısını belirtin."},
{"from": "gpt", "value":… See the full description on the dataset page: https://huggingface.co/datasets/berhaan/clevr-tr.clevr-math-deduplicatedspatial_clevr_numbered_1000