OPPORTUNITY Activity Recognition from Wearable Sensors

OPPORTUNITY is a benchmark dataset for human activity recognition using on-body and object sensors. It contains recordings of daily activities performed by multiple subjects in a sensor-rich environment. The dataset was created by researchers and is hosted on the UCI Machine Learning Repository.

Use Cases

• Classify locomotion activities like walking or sitting from accelerometer and gyroscope data.
• Detect manipulative gestures such as opening a door or drinking using data from object-mounted sensors.
• Train sequence models like LSTMs on the multivariate time-series data to predict activity transitions.
• Benchmark multimodal fusion techniques by combining data from body-worn, object, and ambient sensors.

Strengths

• Benchmark dataset from the UCI repository, a trusted source for machine learning data.
• Contains data from multiple sensor modalities including body-worn, object, and ambient sensors.

Limitations

• Specific row count, column details, and recording duration are not provided in the input metadata.
• Potential for class imbalance or specific subject demographics is unknown without the full dataset documentation.

Provenance

Source

UCI Machine Learning Repository

Collection Method

Data gathered from subjects performing scripted daily activities in a sensor-instrumented environment.