1,082 datasets found

The SBD dataset is a benchmark for semantic edge detection, containing 11,355 images with corresponding labeled semantic edge maps for 20 Pascal VOC classes.

The dataset used in the paper for uncalibrated photometric stereo using fewer images.

The dataset used in the paper for stereo matching, which is a type of computer vision problem.

The dataset used in the paper for color segmentation, which is a type of computer vision problem.

The dataset used in the paper for graph matching problems, which is a type of computer vision problem.

The view planning problem (VPP) for 3D models. The dataset consists of 3D meshes of 20 different objects.

A geometric multi-model fitting dataset

A real-life dataset containing approximately 2100 trajectories of people recorded by a camera taking 5120×3840 resolution images running at 6 frames per second.

Pooling is a simple but widely used layer in modern deep CNN architectures for feature aggregation and extraction. Typical CNN design focuses on the conv layers and activation...

The CGQA-states dataset is a dataset that includes subsets of the CGQA dataset that specifically relate to object states.

The OSDD dataset is a dataset tailored for state detection.

The proposed method for Object-agnostic State Classification (OaSC) combines structured knowledge on object states with visual information that relates the appearance of certain...

CRFs have been applied to a variety of domains, including text processing, computer vision, and bioinformatics.

The dataset used in the paper is a binary matrix, and the authors used it to test their algorithm for Boolean Matrix Factorization (BMF) and Fp-Matrix Factorization (Fp-MF).

The dataset used for testing the gravity-based height estimation method.

The COCO2017 dataset is utilized to evaluate the quantization performance for object detection task.

Human pose estimation has achieved significant progress in recent years. However, most of the recent methods focus on improving accuracy using complicated models and ignoring...

Dataset for hand tremor amplitude measurement using smartphone videos

We create the long-tailed SVHN-LT dataset with maximally 1,000 images per class (head class), and vary ρ from 10 to 100 for different long-tailed versions.

We create the long-tailed CIFAR versions following [7,11] with controllable degrees of data imbalance, and keep the test set unchanged.