21 datasets found

The dataset is used for open-domain human-machine conversation, where the goal is to generate responses to context.

The dataset for the paper AniDress: Animatable Loose-Dressed Avatars from Sparse Views Using Garment Rigging Models

The Topical-Chat dataset is a knowledge-grounded open-domain conversational dataset, which consists of dialogues between two Mechanical Turk workers (a.k.a. Turkers).

Hand segmentation for hand-object interaction using only a depth map

The WISDM dataset was collected using Android based smartphones. The data captured with the embedded tri-axial accelerometer was recorded while the users performed 6 different...

The Distress Analysis Interview Corpus of Human and Computer Interviews (DAIC-WoZ) is a dataset of human-computer interviews for assessing psychological distress.

The DICES-350 dataset is a curated sample of 8k multi-turn conversation corpus generated by human agents interacting with a generative AI-chatbot (Thoppilan et al., 2022) in an...

The HCI gesture recognition dataset consists of a user performing 5 different gestures using the right arm.

The dataset used in the paper for human-art dataset

RF-Capture: Capturing the Human Figure Through a Wall

Human Activity Recognition (HAR) datasets used in the paper for Federated Learning and Federated Unlearning.

The GazeHOI dataset is a comprehensive resource for studying the complex interaction between human hands, objects, and gaze.

VideoAttentionTarget is a video-based gaze target dataset comprising 71,666 frames from 1,331 clips.

GazeFollow is a large-scale dataset consisting of 122,143 images with 130,339 annotations on head-target instances.

Gaze target detection aims to directly associate individuals and their gaze targets within a single image or across multiple video frames.

The authors used the OpenAssistant dataset to construct evaluation datasets for their attacks.

Dataset for tracking faces, hands, and bodies in videos.

Real-world dataset for reconstructing 3D people models from monocular video.

Real-world datasets for reconstructing human avatars from monocular video, including ZJU-MoCap and People-Snapshot.

The authors propose a novel approach that incorporates parameter-efficient tuning to better optimize control tokens, thus benefitting controllable generation.