(pple stealing from the shop) using ML approach. I am thinking of getting images that capture scene like that then build a classifier around it.
Is this a suitable approach? Or anyone with a suitable approach should please help out.
Thanks
I'm not an ML expert by any means, but I think this would be difficult given that shoplifting implies hiding the stolen items (unless you're going to use special kind of imaging; like Infrared Imaging). However, I think another approach would be the movement pattern of a shoplifter. By building a map/path of the shoplifter inside the store, you might be able to find a pattern in the way they move around before stealing.
How about activity recognition?
I agree with @YazanWael. It would be challenging to build a classifier based purely off of image recognition since the only way I can see it being deployed is if you compare the before and after images of the place the object was kept at and the person and their cart/basket to see if you can find the object in either places. However, this model wouldn't work in most of Europe because people often place things in the bags or backpacks they bring along with themselves and it might look shady at first but they eventually do honestly pay for it all . Last, I would argue that setting this up with its nominal accuracy would be much more expensive than just tracking the object codes and people. Facial recognition might be an ethical and legal concern for both the methods if you would like to identify the shop lifter. If not, then the latter is definitely easier to install and cheaper than using image recognition
Обсуждают сегодня