Thousands of Pics (of Millions of Calories) Uploaded Daily to Gimme VMS

I had never seen so many Calories until I started reviewing what Gimme Field and Gimme AutoDrive captures. Every day our users submit over 2,500 pictures of vending machines, micro markets, and other food points of sale. Our system collects and uses these photos to train our AI to detect things automatically. Every image makes the system better and more accurate — and we gather a lot of images.

As an aside, In the context of Calories, the number of consumer packaged food products in a typical picture ranges from 8 to 100, with a median around 50. With packages commonly containing 100 Calorie snacks, that’s about 12.5 million Calories photographed a day on Gimme! Of course, there are exceptions, like the first time I saw a single-wide snack vendor, shown here in the picture.

These labeled images show what we’re training our AI to detect. The long-term value in the insights these photos provide is immeasurable at scale for an operation. By submitting photos with each visit, our system is learning to correlate the “facts” about each visit with how its represented in the picture. This includes information like the number of stock-outs, the arrangement of products, the fill level, the compliance with product swaps, and the general condition of the point of sale. Every visit and every picture makes it better and improves the overall insight into your operation.

We also moved those photos to prominent positions throughout Gimme VMS and many of our customers love being able to review it themselves. They get a first-hand view into the service quality all their accounts are receiving, how their team is merchandising, and if processes are followed, like for cash handling. Notice where the cash was placed in this picture.

Using images to train computer systems is becoming popular because of its application in training self-driving cars. Using an artificial neural network simulates how our own human brain "learns" through processing and storing information. When we spend more time with an idea and practice using it, the more we learn. Similarly, neural networks have self-learning capabilities that improve as more data is made available.

The rate our system is learning is accelerating.

Last year we made the conscious decision to require pictures from every completed vending machine or micro-market service visit. Today we receive 10,000+ images a week from our existing customers. Now that we are opening up Gimme VMS sales to new customers, it means new drivers, new points of sale, and more pictures every day than the one before it. The rate our system is learning is accelerating.  

This is worthwhile news for our customers and prospects. The faster Gimme's neural networks learn and develop, the larger our customer's technological lead grows over their own competitors using static legacy systems.