Annotating a large body of images quickly, accurately and inexpensively would be a valuable capability in scientific, medical and other commercial applications. Machine vision is making progress in these arenas. However, accuracy is not yet sufficient for many applications. In recent years, a complementary solution has become available: crowdsourcing, that is, dynamically recruiting thousands of people to carry out an assigned task from their computer. The team's research suggests that it is possible to combine the complementary strengths of human annotators and machines into a hybrid system that is flexible, accurate, fast and inexpensive. To demonstrate effectiveness and potential commercial opportunity, the team will develop a prototype and a business model around this approach.
As imaging becomes more available and storage inexpensive, the amount of image data will continue to increase. This is true for the scientific, research, geospatial information systems and consumer markets. The proposed effort will address the need to scale annotation and analysis of this data while keeping the process as inexpensive and fast as possible with today's computational power. By combining computer vision and machine learning automations with humans (both experts and non-expert annotators), the system promises to be quickly configurable and trainable across virtually any image analysis challenge.
The award covered the cost of attending Steve Blank's entrepreneurship class organized by the National Science Foundation. This included traveling to Stanford University to attend classes and participate in class-related activities, developing a canvas-based business model, developing a minimum viable product, holding about 120 in-person interviews with potential customers. Three people participated: the technical lead (B. Babenko), the industrial mentor (E. diBernardo) and the PI from Caltech (P. Perona). The team was called `Anchovi'. Anchovi's technology was based on trainable computer vision systems developed in Perona's lab at Caltech and on techniques for crowdsourcing visual tasks also developed by Perona and collaborators at Caltech. The grantees initially had a plan to develop software to allow scientists and commercial entities to train a computer system to annotate images automatically (e.g. detect all cancerous cells in a microscope slide, count all cars in a parking lot from satellite imagery). After developing an MVP and carrying out about 60 interviews the grantees decided that they had not found an `earlyvangelist' (cfr S. Blank's book) and therefore they decided to `pivot' and develop a new business plan. Their new chosen direction was helping consumers organize (store, search, classify) their smart phone pictures in the cloud. After carrying out about 60 interviews it was clear that this was a good product idea; the team developed an MVP as well as two hypotheses for how to acquire customers: directly through the web and by building apps for cloud storage companies (Amazon, Dropbox, ...). At the end of the iCorps period, after a meeting with Dropbox, the Anchovi team was offered to merge their efforts with Dropbox. The deal made sense due to Dropbox's fast-expanding base of customers and plan to become a leading provider of cloud storage for smart phone pictures. In October of 2012 Anchovi was purchased by Dropbox.
