With interest in artificial intelligence (AI) services on the rise, image recognition services promise to drive new capabilities and efficiencies while transforming customer service across diverse industries. Marketers of these services have already pitched them for a wide range of imaginative uses:
- As people enter an upscale retail store, image recognition services could immediately identify high-value customers and gauge their moods.
- The state highway department could quickly review images of thousands of miles of roads to identify potholes and prioritize repairs.
- Manufacturers might assess photographs of the current inventory and rapidly determine what needs to be re-ordered.
- Law enforcement agencies could measure crowd sizes and identify people believed to present security risks.
- Customers of an e-commerce company could take pictures of a product and submit them to the retailer, whose app could locate similar products and allow the customers to purchase them immediately.
Given these and other dramatic possibilities, image recognition, a branch of artificial intelligence and machine learning, is rapidly gaining the attention of businesses and government agencies. But are the image recognition services that are commercially available today capable of delivering on their promises?
To cut through the hype and understand the utility these services truly offer, CapTech tested the six leading image recognition services: Amazon Rekognition, Microsoft Azure Computer Vision, Clarifai, CloudSight, Google Cloud Vision, and IBM Watson. This paper summarizes the results of the tests.
We presented each service with the same set of approximately 4,800 images, distorting many by blurring, overexposing or underexposing, positioning the images at odd angles, and otherwise recreating real-world conditions. We evaluated them across performance and the services’ confidence in nine distinct areas of function including adult content detection, facial detection, facial recognition, mood analysis, text recognition, logo recognition, branded product identification, item classification, and item recognition. We also evaluated overall correctness and average response time of the services.
We found that no one service has a clear lead across all tested functions. With no one-size-fits-all solution available on the market today, we recommend that organizations looking to adopt image recognition be prepared to use multiple vendors to accomplish their strategic goals and objectives. In addition, because of the rapid pace of change in this industry, we recommend that integration of existing systems with image recognition services be architected to provide maximum flexibility so that organizations can switch vendors as needed and adapt to the rapidly changing image recognition landscape.
View Infographic: Help or Hype: An Unbiased Image Recognition Services Vendor Assessment
Read the full research study here.