Extending or correcting the classifications

I see the array ‘selected_items[]’

I also see all of the labels in /usr/share/ti/examples/tidl/classification/imagenet.txt which appears to be a list of all possible items for the classifier (my guess).

selected_items[n] = integer_value

What does this integer_value represent? Probably some result from a classifier???

selected_items[0] = 429; /* baseball /
selected_items[1] = 837; /
sunglasses /
selected_items[2] = 504; /
coffee_mug /
selected_items[3] = 441; /
beer_glass /
selected_items[4] = 898; /
water_bottle /
selected_items[5] = 931; /
bagel /
selected_items[6] = 531; /
digital_watch /
selected_items[7] = 487; /
cellular_telephone /
selected_items[8] = 722; /
ping-pong_ball /
selected_items[9] = 720; /
pill_bottle */

I just want to understand this so that I can see how to extend the classification to other items.
Say I wanted to add a ‘golf_ball’ for the classifier… Does anybody have anything to explain how to add it?

Screen Shot 2019-12-10 at 4.37.50 PM.png

…or when you put a golf_ball in view and it indicates ‘ping-pong_ball’ Any recommendations to improve the classification?

Screen Shot 2019-12-10 at 4.53.30 PM.jpg

I still need to dig through the code and docs to figure this out.

My first goal is to add a ‘golf_ball’ to be recognized. It’s my “hello world” for AI.

With a brief look at Sean Miller’s project, it turned out to be quite simple:

Recognizing golf_ball

I also added a ‘banana’ which it recognized too.

Screen Shot 2019-12-10 at 6.41.13 PM.jpg