AI is amazing. The use cases are everywhere from casual chatting to vision tasks. Take the SAM3 model developed by Meta for instance. This model can create masks on an image or video, which doesn’t sound too amazing until you think about the possible use cases. SAM3, in particular, is used to generate masks which goes hand-in-hand with image segmentation. Mask generation determines which pixels most resemble a text based prompt, where image segmentation gives an overview of the dominant objects in a photo or video. The team that created the SAM3 model is very talented!

Think about the possibilities! A computer can determine if your home is being invaded by an intruder, a computer can determine if your 3D printer is turning into a pile of goo, a computer can determine the bounding area of landmarks and features from satellite images, a computer can determine if you have a cavity from pictures and x-ray results, etc. The list goes on and on!

For example, lets say a civil engineer needs to create a layout of a neighborhood that was built years ago and no CAD models exist. SAM3 can be used to create masks within the satellite image of the neighborhood with certain objects such as roadways, houses, playgrounds, trees, etc. Then, the outline of those masks can be appropriately scaled and imported into CAD as a DXF. Obviously, this CAD should not be used for highly accurate measurements; however, the workflow is very fast and versatile, allowing multiple objects to be measured and imported into a CAD document quickly. Image distortion and image/mask resolution are concerns here, but for quickly selling a client on an idea? Now that is where the power lies!

Imagine you are a farmer and you want to estimate the size of your crop yield from year to year. SAM3 is perfect for this! Imagine you have an ecommerce website that allows your visitors to try on your clothes and accessories without having to visit the store! Mask generation is perfect for this too! Imagine you are a quality control engineer. SAM3 can be used for defect detection along with many other tasks. So, next time you hear from the usual AI hater, know that AI is here to stay and the use cases are endless.

Contact Inontek Solutions to implement Image Segmentation or Mask Generation into your next project!

Lets look at an example to see how good SAM3 is. Below is a satellite image of a random neighborhood. (All Photos are from Google Satellite) First, we will run the SAM3 model to mask all roads within the image…



google_satellite_of_neighborhood

google_satellite_of_neighborhood_masked2

Not bad. The red indicates where the SAM3 model determined there was a “roadway” in the satellite image. Now, lets tell the model to mask all houses within the image…



google_satellite_of_neighborhood_masked3

Will you look at that! SAM3 is amazing. All within a couple of seconds, AI has determined exactly where the roads and houses are within the initial satellite image. Thank you for reading this blog post, we hope to see you again in the future and remember to tell your network about Inontek Solutions!