Wednesday, January 29, 2014

competition for energy-efficient computing (image recognition)

This figure shows the idea of the competition. The inputs will be a set of images. The score is the ratio of the accuracy and the amount of energy consumed. The participants have the freedom to offload computation.

The competition will be held in 2015 with a major conference (one candidate is IEEE International Symposium on Circuits and Systems in June). The winner must be able to recognize or classify at least half of the images (i.e. the accuracy must exceed 50%).
  • first prize is a cash award of USD$5,000
  • second prize $3,000
  • third $2,000. 
The participants will receive a set of training data at least one month in advance.


  1. Some questions about how this will work...

    Will we specify what hardware they will use?
    If so, will there be funding or hardware to be donated?
    If not, how do you compare power/energy results? Or, is that part of the competition?

    NVIDIA can probably provide some hardware for prizes, but not hardware platforms for everyone to use!

  2. This seems like a great idea! There are a wide range of image recognition tasks that might be appropriate for this kind of evaluation. Do you have ideas about the type of recognition? Or about the range of power to consider, 100s of watts? There were some neat demos of multi-class detection (output is a bounding box and label) of 200 classes of objects on a cell phone at our ImageNet Challenge workshop at ICCV last year. I think this was running at ~0.5-1 frames per second.


  3. Very interesting article! By the way you are running a great blog. Thanks for sharing.
    secure virtual data room