Amazon Web Services has announced a new hardware product called AWS Panorama Appliance, a device that adds machine learning capabilities to any network-connected camera. The company is pitching the technology for industrial uses, making it easier to monitor footage from cameras in facilities and detect issues that need to be corrected, like factory workers not wearing face masks.
Plug and play — Computer vision trained to detect objects and activity in video is nothing new — Amazon's own Rekognition software can be used to analyze photo and video. The difference with Panorama is that the machine learning models can be hosted locally on the Panorama Appliance and run against video feeds in real-time so that response times are fast, and data never has to leave the premises. The cameras don't even need to be connected to the internet.
With the Panorama Appliance connected to the same network as the IP cameras, the results of any machine learning analysis can be displayed on a monitor like one in a building's security room. Organizations don't have to buy new cameras or redo their CCTV system — that simplicity is what Amazon is selling, along with the fact that some organizations may have data governance rules that prohibit data from leaving their premises.
Panopticon — Advancements in machine learning for video have been controversial to say the least, particularly as law enforcement has used the technology to identify suspects in crimes, sometimes leading to false positive arrests. But retail businesses have stepped up their usage as well in hopes of quickly identifying suspicious activity. Instances of misclassifying people of color, and a high potential for abuse, have led critics including civil rights advocates to call for slower roll-out of new video surveillance. Amazon itself said it would temporarily stop offering Rekognition to law enforcement agencies while it reevaluates the program.
The company is really emphasizing more industrial use cases for Panorama, but that still may draw pushback as workers complain about having their performance closely monitored. In Amazon fulfillment centers, employees working as "pickers" carry handheld computers that monitor their speed retrieving items, and warn them if they're falling behind expectations. That way of assessing workers purely based on granular data is considered cold and demeaning, and can create anxiety. Nobody wants to have someone watching over their shoulder the whole time they work.
Amazon is constantly trying out new technology inside its warehouses to increase speed and productivity, but as it does so, its injury rate has actually increased to higher than the national average.