The sensor will give temperature data of each pixel of the 8 x 8 array in a 2-byte format. The pixels are infrared sensors which measure an averaged temperature value that each pixel sees within their respective fields of view. The full field of view of the entire sensor array is about 60 degrees both vertically and horizontally, and each pixel’s field of view will be roughly 7.5 degrees horizontally and vertically. If a person were to stand in the field of view of one or more pixels, then those pixels will measure a higher temperature than the surrounding pixels, assuming the background temperature was cooler than the person. A host device, such as an Arduino or Raspberry Pi, will send the appropriate command to the slave Grid-Eye device under I2C control, and the Grid-Eye will send out the pixel data for all 64 pixels back to the host under I2C control from the host. It takes a total of 135 bytes to receive a full frame of data.
Since you provide the code for the host, you can do whatever you want with that data, including assign a color to each temperature value (typically red is assigned to higher temperatures, but that’s up to you). You could choose to create an 8 x 8 image using those color assignments and display that on whatever display is compatible with the host device, or transmit that to a PC if your host is capable of doing that.
If you get the AMG8834EVAL kit and use the software they provide, the output of the kit’s host (an Atmel/Microchip ATSAMD21) can be sent to a PC and displayed within a GUI. As for your question about the three Levels, according to what I read, all three levels are included in the microcontroller firmware, and you just select which you want to use. See the video below for a demo of the Panasonic eval board:
Regarding your questions about the Adafruit Grid-Eye with Raspberry Pi, you will have to decide whether you convert the temperature data to video or not, depending on what you are trying to do. If you intend to display the field of view, you will obviously have to create an image at some point, and I would assume most image processing algorithms require an array of data to work with.
Using Pygame is not a requirement with the Raspberry Pi, but it might be a useful way to generate an image. I am not familiar with that software. I did find a link to a project here were someone used the Grid-Eye and a Pi, and fed the output to a PC. I have not studied it in detail, but he gives the code, so it may well help you along in that regard.
Here’s one additional example, using the Sparkfun SEN-14607, which is another board we sell with a Grid-Eye mounted on it. They give a number of basic examples, and then give an example using “Processing” which is a programming language often used to create graphics. They show an image generated with this which is quite nice.