Workshop on Advancing Computer Vision with Humans in the Loop

There are several venues in which development of machine algorithms to analyze images can benefit from human involvement. Beginning to understand the human visual recognition system can provide valuable insights for advancing machine vision. Humans can be involved in the task of collecting and labeling data, especially with the increasing popularity of tools like LabelMe and Amazon Mechanical Turk. Human studies can be performed to evaluate algorithms for super-resolution, segmentation, matting, etc. There are perhaps other avenues to exploit human involvement that have not been explored yet. This workshop provides a forum for works that actively keep humans in the loop for advancing computer vision.
Call for Papers
Papers must describe high-quality, original, and novel research. Areas of interest include all areas where human interaction is exploited to advance computer vision through improved algorithm design, data collection, evaluation, etc.
Some specific areas of interest include, but are not limited to:
* Understanding the human recognition system to advance computer vision;
* The use of Amazon Mechanical Turk and other online labeling tools: which tasks are appropriate and which ones are not, how quality is maintained;
* The benefits, biases, limitations and trade-offs involved with the use of human vocabulary to query and define dataset labels;
* The benefits, biases, limitations and trade-offs involved with the use of human subjects for labeling ground truth in images;
* Active learning geared towards human aspects e.g. accounting for varying effort and costs involved for different label types;
* The use of human subjects in evaluating vision algorithms such as super-resolution, matting, segmentation, or cut-outs; and
* Other creative venues to leverage human input to advance computer vision that have not been explored in the community thus far