Generative AI

How we use Generative AI to quickly deploy Computer Vision in factories & warehouses

Quickly go from pilot to full deployment using Generative AI.

Nov 28, 2022

Generative AI
Generative AI

Before we get into the specifics of how the Logistify Platform uses Generative AI to quickly deploy AI models in factories and warehouses, it’s useful to understand what this technology is and get a sense of just how popular it is at the moment (as of this writing in November 2022).

The most consequential shift in the internet has just happened. We now have cheap, high-quality, easily-accessible & fast AI models for generating original and realistic images, videos, text, music and more. This technology is called Generative AI. Popular examples of Generative AI models include Stable Diffusion, Open AI GPT-3 & DALLE-2.

Real world applications of Generative AI are being realized currently by startups at such a fast pace that it has already been validated by real revenues.

In the developer community, Generative AI is seeing the fastest adoption we’ve ever seen. For example, Stable Diffusion easily tops the trending charts of GitHub repositories by a wide margin. Its growth is far ahead of any recent technology in infrastructure or crypto (see the figure below from a16z blog).


How Generative AI can accelerate AI deployment in factories and warehouses

It's a well known fact that the supply-chain industry has the slowest adoption of technology, more so in warehouses and factories. Not only are the sales cycles so long, but also the time to deployment of a technology & experiencing its value is frustratingly long. For computer vision models, the time to deployment involves, among many steps, collecting good quality & consistent training data.

Generative AI can help reduce the time to deployment of Computer Vision in factories and warehouses. By simply generating realistic images of the customer's SKU items in different scenarios/environments (e.g. poor lighting, obstructions, etc) & with different types of defects (e.g. damages, wear & tear, etc), we can have initial training data in seconds. See examples below from the Logistify Platform.

FIGURE 1: Example 1 - A factory for baking flours. Realistic images of bales of baking flour with different defects are generated as initial training data.

FIGURE 2: Example 2 - A factory for steel pipes. Realistic images of a specific grade of steel pipes are generated as initial training data.

We can then train a Computer Vision detection model, and deploy this AI model before the customer even picks up their phone to take pictures for the pilot. The base model is then further tuned by annotating a small sample of their data, using a Data-Centric AI approach, thereby increasing accuracy levels significantly.

Figure 3: Additional image data is annotated and used to train the base model

Labelled Maize Meal


Figure 4: The video captures of the operations in the Logistify Platform


Figure 5: Final output of the trained inventory-detection-and-counting AI model used during receiving of inventory. This reduces the number of human checkers at the receiving dock thereby reducing labor costs.


So why is this important for us as a startup?

Speed is everything in startups.

  • Generative AI enables us to have a huge repository of pre-trained, out-of-the-box AI models for different inventory (e.g. Hardware, Food & Beverage, etc), inventory conditions (e.g. defected, damaged, etc) and factory/warehouse environments (e.g. dark lighting, etc). When a new customer signs up, our in-App onboarding questionnaire figures out which pre-trained model to allocate them. After onboarding, the base model is then further tuned by annotating a small sample of their data.

  • The faster we can show value to the customer, the faster we can close Enterprise-Sales conversations with the decision maker for a long-term engagement (Typically multi-year agreements).

  • It democratizes the process of data preparation. The simplicity of the technology enables Factory/Warehouse inventory controllers to write their own text descriptions and generate their own data for training AI models. They are the domain experts and understand their inventory better than us.

The ease-of-use of this technology enables us to scale much faster in an industry that is traditionally very slow to technology adoption.


Request Access Today!

If you’d like more information about the Logistify Platform, reach out to us for a detailed demo and discussion here.

GET STARTED TODAY

Reduce inventory losses & labor costs