The CUHK-X Multimodal Human Activity Challenge is an international competition organized by the AIoT Lab at The Chinese University of Hong Kong. It invites teams from all over the world to advance privacy-preserving human activity recognition, understanding, and reasoning using non-RGB sensing modalities — depth, IMU, millimeter-wave radar, skeleton, thermal, and infrared — built on the CUHK-X benchmark.

Teams first compete online on Kaggle across two parallel tracks. Shortlisted teams then enter a post-freeze verification and private-data evaluation process conducted by the Organizing Committee, and the finalist teams advance to the Grand Finals at UbiComp 2026 in Shanghai for technical presentations, Q&A, and awards.

A Two-Track Challenge

The CUHK-X Challenge is organized as two parallel tracks, each with its own Kaggle competition, leaderboard, and USD $10,000 prize pool:

• Small Model Track — Human Activity Recognition (HAR): build lightweight multimodal models that classify 40 daily activities from depth, IMU, mmWave radar, and skeleton data.
• Large Model Track — VQA for Action Understanding and Reasoning: apply multimodal large models to visual question answering on privacy-preserving videos, covering human action understanding (HAU) and human action reasoning (HARn).

How to Participate

1. Register your team on the official challenge website: openaiotlab.github.io/CUHK-X-Challenge. Registration on the official website is required to be eligible for prizes, announcements, and finals invitations.
2. Join your track(s) on Kaggle using exactly the same team name as registered — Small Model Track (HAR) · Large Model Track (VQA). Teams may enter one or both tracks; certificates and shortlist notifications are matched by team name.
3. Download the data, train your model, and submit predictions on Kaggle before the leaderboard freeze at 23:59 UTC on September 15, 2026 (07:59 UTC+8 on September 16).
4. Advance to the verification and evaluation stage: the Top 15 teams per track on the Kaggle private leaderboard are invited to submit verification materials by 23:59 UTC on September 18, 2026 (07:59 UTC+8 on September 19). The Organizing Committee uses the submitted code, model artifacts, configuration, and setup instructions to reproduce each solution and conduct the private-data evaluation. The Top 6 teams per track, subject to verification, advance to the Grand Finals.
5. Compete in the Grand Finals: finalists participate at UbiComp 2026 in Shanghai on October 12, 2026, give a 5-minute technical presentation in English followed by 5 minutes of Q&A, and attend the awards ceremonies. There is no additional on-site private-test inference or reproduction round during the Grand Finals.

Awards

Each track carries an independent prize pool of USD $10,000 (USD $20,000 in total). The following prizes apply independently to both the Small Model Track and the Large Model Track:

• 1st Place: USD $6,000
• 2nd Place: USD $3,000
• 3rd Place: USD $1,000

Three special awards will additionally be presented per track (prize amounts to be announced):

• Best Report Award: open to top-ranked teams in each track that receive a formal invitation from the Organizing Committee; recipients are selected by the review committee.
• Most Popular Award: determined by an online community vote, opening once the finalists are announced on October 1 and closing at 23:59 UTC on October 10, 2026 (07:59 UTC+8 on October 11).
• Best Faculty Advisor Award: selected by the competition committee in recognition of outstanding mentorship, considering factors such as team performance and overall contribution to the challenge.

Beyond the cash prizes, every participating team is recognized through a five-tier certificate system (per track). Tiers are nested — each team receives only its highest-qualifying award:

All certificates are issued electronically to the email address provided during team registration.

Travel grants: every finalist team (Top 6 per track) attending UbiComp 2026 in person receives a travel grant of up to USD $500, reimbursed against actual expenses. Teams unable to travel may join the finals remotely via Zoom and remain eligible for all prizes and awards; travel grants apply only to in-person attendance.

Summary of Key Dates

All deadlines are governed by UTC; the Hong Kong / Shanghai equivalent (UTC+8) is shown in brackets. The Grand Finals programme is given in Shanghai local time (UTC+8).

• June 20, 2026: Competition launch — both Kaggle competitions open for registration and submissions; the challenge dataset is publicly released. (already started)
• September 15, 2026, 23:59 UTC (September 16, 07:59 UTC+8): Kaggle leaderboard freeze. The Top 15 teams per track are notified and invited to the verification and evaluation stage.
• September 18, 2026, 23:59 UTC (September 19, 07:59 UTC+8): Submission package due from shortlisted teams.
• September 19–30, 2026: Verification and evaluation stage — committee-run code and reproducibility verification and private-data evaluation.
• October 1, 2026: Final Top 6 teams per track announced. Technical reports from formally invited teams are due at 23:59 UTC on October 1, 2026 (07:59 UTC+8 on October 2).
• October 12, 2026: Grand Finals at UbiComp 2026 in Shanghai — finalist presentations, Q&A, and awards ceremonies.

Competition Process

The challenge consists of two independent parallel tracks, each with its own Kaggle competition, leaderboard, prize pool, and evaluation criteria. Teams may participate in one or both tracks. Both tracks follow the same three-part process.

Kaggle Open Competition (June 20 – September 15, 2026 · 3 months)

Teams register on the official challenge website, join the corresponding Kaggle competition with the same team name, download the released challenge data, train their models, and submit prediction CSV files on Kaggle. The leaderboard freezes at 23:59 UTC on September 15, 2026 (07:59 UTC+8 on September 16), after which the Kaggle private leaderboard determines the Top 15 shortlisted teams per track.

Verification and Evaluation (September 19 – September 30, 2026 · 2 weeks)

Shortlisted teams must submit their verification package by 23:59 UTC on September 18, 2026 (07:59 UTC+8 on September 19). The package includes full training and inference code, final model weights or API configuration as applicable, a single-entry inference script, reproduction README, final submission information, and a signed honor declaration.

The Organizing Committee will run each submitted solution using the team’s submitted code, model artifacts, configuration, and documented setup instructions. This verifies that the solution executes successfully and produces results consistent with the team’s selected Kaggle submission. If a setup or execution issue arises, the committee may contact the team privately for clarification or a closed verification session. A team whose submitted solution cannot be verified may be replaced by a reserve team according to private-leaderboard order.

During the same stage, the committee will evaluate each submitted solution on organizer-held private data. The resulting score contributes 30% of the final score for finalist teams. There is no additional private-test inference during the Grand Finals.

Technical Report and Best Report Award

Following the leaderboard freeze, a broader group of top-ranked teams in each track will be formally invited by the Organizing Committee to submit a technical report for consideration for the Best Report Award. Reports from invited teams are due by 23:59 UTC on October 1, 2026 (07:59 UTC+8 on October 2).

All invited teams remain eligible for the Best Report Award whether or not they advance to the Grand Finals. Award recipients who do not advance will be notified by email and receive an electronic certificate. For finalist teams, the report also contributes to the 20% Technical Report component of final scoring. Selection for the Best Report Award is conducted independently of Grand Finals qualification.

Conference Presentation (October 12, 2026 · Shanghai · 1 day)

The Top 6 teams per track are invited to the Grand Finals at UbiComp 2026 in Shanghai. Each team gives a 5-minute technical presentation in English followed by 5 minutes of Q&A with the jury. The Grand Finals focus on presentations, technical discussion, judging, and award ceremonies. Teams unable to travel may participate remotely via Zoom and remain eligible for all prizes and awards.

Tracks and Final Scoring

Small Model Track — Lightweight Human Activity Recognition

This track targets resource-constrained edge deployment in smart-home and healthcare scenarios — applications such as Alzheimer’s monitoring, fall detection, and elderly care, where models must run on low-power devices with limited memory and compute. Participants build lightweight multimodal models that fuse depth imagery, IMU streams, mmWave radar, and skeleton keypoints to classify 40 daily activities under a strict cross-subject evaluation protocol.

• Task: 40-class human activity recognition (cross-subject)
• Cross-subject split: training on users 1–9 and 16–24; testing on users 10–11 and 25–26
• Modalities: depth, IMU, mmWave radar, skeleton, infrared, thermal
• Model constraints: conventional architectures (CNN / RNN / Transformer); model size ≤ 100 MB; no large pretrained backbones
• Prize pool: USD $10,000

Large Model Track — Multimodal VQA (HAU & HARn)

This track pushes the frontier of large vision-language models on non-RGB modalities. Participants tackle human action understanding (HAU) and human action reasoning (HARn) through visual question answering on privacy-preserving videos. There is no parameter limit, encouraging exploration of prompt design, modality alignment, and fine-tuning at scale.

• Task: visual question answering on privacy-preserving videos (HAU and HARn)
• Cross-subject split: training on users 1–9 and 16–24; testing on users 10–11 and 25–26
• Modalities: depth, thermal, infrared, skeleton, IMU, and mmWave radar
• Model constraints: no parameter limit; large vision-language models encouraged
• Prize pool: USD $10,000

Final Scoring (Finalists)

The final ranking of the Top 6 teams in each track is determined by:

• Kaggle private leaderboard — 20%
• Organizer-run private-data evaluation — 30%
• Reproducibility (selection stage) — 10%
• Technical report — 20%
• Presentation — 10%
• Model efficiency — 10%

The private-data evaluation is conducted by the Organizing Committee during the post-freeze verification period using organizer-controlled private data and the official metric for each track. A detailed scoring rubric is published separately before the finals.

The CUHK-X Dataset

Most large vision-language models still depend almost entirely on RGB data, while modalities such as depth, thermal imaging, IMU, and millimeter-wave radar remain severely underrepresented — largely due to the lack of large-scale, high-quality paired multimodal datasets.

CUHK-X, built by the AIoT Lab at CUHK, addresses this gap with 64,267 fully synchronized samples across seven modalities (RGB, depth, thermal, infrared, skeleton, IMU ×5, and mmWave radar), collected from 30 participants performing 40 daily activities in two real-world indoor environments. Annotations follow a Ground-Truth-First strategy that combines LLM-generated scene descriptions with human review to ensure temporal and logical consistency. The dataset supports three progressive tasks: HAR (action classification), HAU (action understanding), and HARn (action reasoning).

RGB data are excluded from the challenge; all remaining modalities are permitted. The cross-subject split assigns 18 participants to training, 4 to the public test set, and 8 to a held-out private test set.

The track-specific challenge data are available from the following mirrors:

• Small Model Track: Hugging Face · Google Drive · Baidu Netdisk
• Large Model Track: Hugging Face · Google Drive · Baidu Netdisk

Further details are available on the CUHK-X dataset homepage and in the GitHub repository.

Eligibility and Team Rules

Eligibility

• Open to students, researchers, and industry teams worldwide
• Cross-institution and cross-country teams are permitted
• Members of the CUHK AIoT Lab and their direct collaborators are ineligible for prizes
• Participants must comply with Kaggle’s terms of service

Team Rules

• Team size: 1–3 members (a faculty advisor is not counted)
• Each individual may join only one team per track; participation in both tracks is allowed
• Team mergers on Kaggle lock 7 days before the submission deadline

Finals Programme

Date: Monday, October 12, 2026 (pre-conference day)
Venue: Yangtze River Hall, 5F, Shanghai International Convention Center (SICC), Shanghai
Language: English · All times are Shanghai local time (UTC+8); UTC equivalents in grey

The Large Model Track runs in the morning and the Small Model Track in the afternoon, each closing with its own award ceremony. Every team gives a 5-minute presentation followed by 5 minutes of Q&A. Finalists in the morning session should be seated by 09:45 (01:45 UTC); those in the afternoon session by 14:15 (06:15 UTC).

Presentation order within each track is drawn on site. All finalists, jury members, and attendees must be registered for UbiComp / ISWC 2026. Remote participants joining via Zoom: the programme runs 01:45–09:20 UTC on October 12, 2026. Times may be adjusted on the day.

Competition Rules

• Fair play (both tracks): manual labeling of test samples, use of test-set ground-truth labels in training (in any form), and multi-account registration or collusion between teams are strictly forbidden.
• Small Model Track: no large pretrained backbones, and no closed-source APIs or LLMs — whether for development or for labeling training data.
• Large Model Track: any pretrained model (including large vision-language models) and closed-source APIs are allowed; LLM-based pseudo-labeling and prompt engineering are encouraged.
• IP and code usage (Kaggle standard): participants retain full copyright on their code and models; all competition data remain the exclusive property of the CUHK AIoT Lab under the dataset license. Finalist teams (Top 6 per track) must open-source their solutions under Apache 2.0 within 30 days of the finals or decline their finalist status; non-finalist code is deleted after the competition.