The Opportunity challenge: A benchmark database for on-body sensor-based activity recognition. Chavarriaga, Sagha, Calatroni, Digumarti, Troster, Millan, Roggen. Pattern Recognition Letters 2013.

1. Human action recognition requires multimodal sensor setups, which leads to interesting ML problems
2. Lack of benchmarks for activity recognition <there are benchmarks – maybe they are no good though>
   1. In many cases, data isn’t shared, or the approach is highly tailored to the specific sensor and environmental setup considered
3. Action recognition requires “tackling issues ranging from the feature selection and classification (…), to decision fusion and fault-tolerance (…).”
4. They recorded “realistic daily life activities in a sensor-rich environment…” <I think no head-cam>
5. They really wire up all the items in the environment
6. Mention some other datasets:
   1. Placelab: Environment where almost everything is wired up, married couple lived there for 15 days.  <Sensors are comprehensive and very hardcore! Even monitor water flow and electricity usage at each plug.>
      1. http://www.mit.edu/~intille/Pervasive05.html
   2. Van Kasteren: 28 day recording, data is much less rich
   3. Darmstadt routine dataset: Just 2 accelerometers in arm, narration
   4. TUM Kitchen <already discussed in blog>
   5. HASC Corpus: Just 1 accelerometer, but on many people
   6. Also there is something called body sensor network contest, but this seems to be devoted to recording very short activity episodes
7. In this dataset, there is a loosely scripted series of activities the subject is instructed to undertake
8. The rest is details on the challenge they propose for their dataset <skipping>