CMI - Detect Behavior with Sensor Data

Overview

This hackathon challenges you to build a model that can distinguish body-focused repetitive behaviors (BFRBs) from everyday gestures using sensor data from a wrist-worn device. You'll work with data including movement, heat, and proximity readings to develop a predictive model. Your work can help improve wearable devices designed to detect BFRBs, supporting mental health treatment.

Requirements

Teams can be any size, but you must accept the competition rules by August 26, 2025, and the team merger deadline is also August 26, 2025. Submissions must be made through Kaggle Notebooks that run within 9 hours on CPU or GPU and have internet access disabled. You are allowed to use publicly available external data and pre-trained models. For each 'sequence_id' in the test set, you need to predict the corresponding 'gesture'.

Prizes

The competition offers a total of $50,000 in prizes. The top six teams will be awarded: 1st place wins $15,000, 2nd place gets $10,000, 3rd place receives $8,000, 4th place is awarded $7,000, and 5th and 6th place each get $5,000.