Object classification system and/or methods

A method for context-specific classification of containers along a conveyor line in an assembly context, the assembly context associated with a container appearance and a target ingredient for insertion into the containers, the context-specific classification comprising a binary classification of target ingredient presence within a respective container, the method comprising:

• generating a set of context-specific training data for the assembly context by automatically labeling segmented image regions of containers with the binary classification according to a predetermined pattern;
• with the set of context-specific training data, training a context-specific classification model to predict the binary container state for containers in the assembly context;
• with the context-specific classification model, determining a set of binary container state predictions for identified containers in the assembly context;
• and based on the binary container state predictions, controlling a robot to insert the target ingredient into at least a subset of the identified containers, wherein the context-specific classification model comprises:
• a first neural network model comprising a first convolutional neural network (CNN) backbone with a first classification head;
• a second neural network model comprising a second CNN backbone with a second classification head;
• and a third neural network model, wherein training the context-specific classification model comprises updating weights of the first CNN backbone, first classification head, and second classification head based on the context-specific training data.