Skip to main content

Ego4D and EgoExo4D Challenge 2024

Overview​

At CVPR 2024, we will host 16 challenges including 2 new challenges (Goal Step and Ego Schema), representing each of Ego4D’s five benchmarks. Included in the 16 challenges hosted at CVPR are two teaser Ego-Exo4D challenges (Ego-Pose Body and Ego-Pose Hands). Please find details below on the challenges:

Ego4D challenges​

Episodic memory:​

  • Visual queries with 2D localization (VQ2D) and Visual Queries 3D localization (VQ3D): Given an egocentric video clip and an image crop depicting the query object, return the most recent occurrence of the object in the input video, in terms of contiguous bounding boxes (2D + temporal localization) or the 3D displacement vector from the camera to the object in the environment.
    • Quickstart: Open in Colab
  • Natural language queries (NLQ): Given a video clip and a query expressed in natural language, localize the temporal window within all the video history where the answer to the question is evident.
    • Quickstart: Open in Colab
  • Moments queries (MQ): Given an egocentric video and an activity name (e.g., a “moment”), localize all instances of that activity in the past video
  • EgoTracks: Given an egocentric video and a visual template of an object, localize the bounding box containing the object in each frame of the video along with a confidence score representing the presence of the object.
  • Goal Step: Given an untrimmed egocentric video, identify the temporal action segment corresponding to a natural language description of the step. Specifically, predict the (start_time, end_time) for a given keystep description.
  • Ego Schema: Given a very long-form video, evaluate the capabilities of modern vision and language systems.

Hands and Objects:​

  • Temporal localization: Given an egocentric video clip, localize temporally the key frames that indicate an object state change.

Audio-Visual Diarization:​

  • Localization and Tracking: Given an egocentric video, recognize each person in a scene and maintain their identity through a long time span.
  • Speech transcription: Given an egocentric video clip, transcribe the speech of each person.

Social Understanding:​

  • Looking at me: Given an egocentric video clip, identify whether someone in the scene is looking at the camera wearer.

  • Talking to me: Given an egocentric video clip, identify whether someone in the scene is talking to the camera wearer.

Forecasting:​

  • Short-term hand object prediction: Given a video clip, predict the next active objects, and, for each of them, predict the next action, and the time to contact.
    • Quickstart: Open in Colab
  • Long-term activity prediction: Given a video clip, the goal is to predict what sequence of activities will happen in the future. For example, after kneading dough, list the actions that the baker will do next.

Other Ego4D challenges which are not part of CVPR 2024 workshop remain open on EvalAI website for submissions but are not eligible for prizes.

Dataset​

Ego4D challenge participants will use Ego4D’s annotated data set of more than 3,670 hours of video data, capturing the daily-life scenarios of more than 900 unique individuals from nine different countries around the world. Unique train, validation and unannotated test sets are available to download per challenge at https://ego4d-data.org/docs/.

This year's challenge we will continue to use Ego4D v2.0 which contains ~2X train and val annotations for Forecasting, Hands & Objects and NLQ, a number of corrections and usability enhancements, and two new related dataset enhancements (Ego Schema and Goal Step). The test set remains the same as previous versions of the challenge. More details can be found here.

EgoExo4D teaser challenges​

Ego-Exo4D is a diverse, large-scale multi-modal multi view video dataset and benchmark challenge. Ego-Exo4D centers around simultaneously-captured ego-centric and exocentric video of skilled human activities (e.g., sports, music, dance, bike repair).

Here are the specific teaser challenge tracks we will host at EgoVis workshop during CVPR 2024.

  • Ego-Pose Body: Given an egocentric video, estimate the 3D body pose of the camera-wearer. Specifically, predict the 3D position of the 17 annotated body joints for each frame.

  • Ego-Pose Hands: Estimate the 3D locations of the defined hand joints for visible hand(s). Specifically, estimate the (x,y,z) coordinates of each joint in the egocentric coordinate frame.

We will be using EgoExo4D dataset for running these challenges. Please find the documentation here for the dataset.

Participation Guidelines​

Participate in the contest by registering on the EvalAI challenge page and creating a team. All participants must register as a part of a “participating team” on EvalAI to ensure the submission limits are honored. Participants will upload their predictions in the format specified for the specific challenge, and will be evaluated on AWS instances by comparing to ground truth predictions. Instructions for training, local evaluation, and online submission are provided at EvalAI. Please refer to the individual EvalAI pages for each challenge for submission guidelines, task specifications, and evaluation criteria.

Dates​

  • Ego4D challenges will launch on March 15, 2024 with the leaderboard closing on May 30, 2024.
  • EgoExo4D challenges will launch on March 29, 2024 with the leaderboard closing on May 30, 2024.
  • Winners for both will be announced at the FIrst Joint Egocentric Vision Workshop at CVPR 2024.

Competition Rules and Prize Information​

Competition rules can be found here. Additionally, we are thrilled that FAIR is able to offer the following prize thresholds for challenges:

  • First place: $1500
  • Second place: $1000
  • Third place: $500

This year, to encourage the exploration of foundational models, additional prizes will be provided to challenge winners whose models can perform on multiple Ego4D tasks. All foundational models will be assessed by the terms and conditions of the rulebook.

Challenge Reports​

In addition to the submission on EvalAI, participants must submit a report describing their method to the workshop CMT (link TBD). In addition to your method and results, please remember to include examples of positive and negative results (limitations) of your model. These validation reports will be evaluated by challenge hosts from the Ego4D consortium before winner determination can be made. Similarly, challenge validation reports, research code from winning entries, and names of participants from the winning teams for all successful submissions must be shared publicly with the research community. More details can be found on the EgoVis workshop page.

Acknowledgements​

The Ego4D challenge would not have been possible without the infrastructure and support of the EvalAI team. Thank you!

Organizers​

  • Selam Mehretu
  • Suyog Jain
  • Andrew Westbury
  • Santhosh Kumar Ramakrishnan
  • Chen Zhao
  • Merey Ramazanova
  • Satwik Kottur
  • Vincent Cartillier
  • Yifei Huang
  • Qichen Fu
  • Hao Jiang
  • Leda Sari
  • Eric Zhongcong Xu
  • Francesco Ragusa
  • Tushar Nagarajan
  • Hao Tang
  • Kevin Liang
  • Weiyao Wang
  • Karttikeya Mangalam
  • Raiymbek Akshulakov
  • Jinxu Zhang
  • Shan Shu
  • Gabriel Pérez Santamaria
  • Juanita Puentes
  • Maria Camila Escobar Palomeque
  • Yale Song
  • Matt Feiszli
  • Giovanni Maria Farinella
  • Mike Z. Shou
  • Pablo Arbelaez
  • Jianbo Shi
  • Kristen Grauman

Past Challenges / Winners​

CVPR Workshop 2023 (June 19, 2023)

ECCV Workshop 2022 (Oct 24, 2022)

CVPR Workshop 2022 (June 19, 2022)