looking for some ideas on how to improve this more its meant to act as the eyes for a seperate AI model i made that this will work with.
#🔒 Looking for ideas
winged arch
opaque sinewBOT
looking for some ideas on how to improve this more its meant to act as the eyes ...
@winged arch
Python help channel opened
wet nexus
use enum instead of a long dictionary with numbers to names
can make confidence a variable which decreases if no matches found but you know there should be one, may not be appropriate in your case
winged arch
anything else like any other ideas associted with like how to mayble make more classes or stuff able to be recognized
winged arch
use enum instead of a long dictionary with numbers to names
explain further
wet nexus
winged arch
k i figure it out i hate using enum
wet nexus
Labels(IntEnum):
Person = 1
Bicycle = auto()
Car = auto()
...
would i put in within the def statement
wet nexus
no
can make BBox a dataclass
so you don't have to use all these bare variables like ymin, xmin, ymax, xmax = bbox
can do a class for the whole thing
winged arch
what is a data class
wet nexus
nicer way to associate variables with a thing
instead of manually like what you're doing there
see how you have to store 2 things in the var name like that
ymax
anyway it's more an evolution to cleaner programming, not going to effect your functionality
winged arch
ok is that associated with OOP ive heard alot about that
wet nexus
not really, I mean technically it is but not so much
it basically just allows you to use less "basic" variables and dicts and stuff
fathom rivet
Well, more about making a "record" like object with names. You can use Python classes for OOP (which is what we largely do) but they are also in a sense records with named fields. Eg:
class BBox:
def __init__(self, ymin, xmin, ymax, xmax):
self.ymin = ymin
self.xmin = xmin
self.ymax = ymax
self.xmax = xmax
then you could go bbox = BBox(*detections['detection_boxes'][i]) and then use bbox.ymin etc to refer to the stored values i.e. using nice names.
wet nexus
allows you to do stuff like
fathom rivet
Dataclasses let you define these more succinctly:
from dataclasses import dataclass
@dataclass
class BBox:
ymin: int
xmin: int
ymax: int
xmax: int
and another way stacking them to have even more granular
@dataclass
class XY:
x: int
y: int
@dataclass
class BBox:
min: XY
max: XY
# ...
# now you can talk to your data like this:
box.min.y
box.max.y
# etc
print(box.min.x) # and so on
Now show the assignment statement to unpack his detections['detection_boxes'][i] into this nested structure 🙂
Anyway, yes, that works though personally I'd invert the nesting for box.y.min.
The OP should use whatever they find more expressive and clear.
winged arch
but would these changes yall are saying i should make allow me to import these functions into the other program with the main ai that im planning on this program being the eyes for
fathom rivet
You can import the functions (and the class names) as you see fit. They're just names in this modules namespace. And import statement just pulls names from another namespace in for use here.
(Well, really, pulling in their values, but the effect is what you expect.)
winged arch
gotcha so basicly just do it as i usually would for anything else
that im importing in that i make
and it should work correctly
?
fathom rivet
Yes.
opaque sinewBOT
Python help channel closed