The retail object reoognition model recognizes retail items detected on a checkout counter by the retail object detection model. This model encodes retail items into embedding vectors and predicts their labels based on the similarity to the embedding vectors in the reference space.
The training algorithm optimizes the network to minimize the embedding output distances (cosine similarity) between the positive images and the anchor image while maximizing the distances between the negative images and the anchor image.
The primary use case intended for these models is to recognize items detected on a checkout counter.
The datasheet for this model is captured in it’s model card host at NGC.