# Rethinking Loss Design for Large-scale 3D Shape Retrieval

## TL;DR

This paper introduces the Collaborative Inner Product Loss (CIP Loss), a novel loss function for 3D shape retrieval that enforces discriminative and well-clustered shape embeddings with clear geometric interpretation, achieving state-of-the-art results.

## Contribution

The paper proposes CIP Loss, a new loss function that improves 3D shape embedding discrimination and clustering, with easy integration and better geometric clarity compared to previous methods.

## Key findings

- Achieved state-of-the-art results on ModelNet and ShapeNetCore 55 datasets.
- Demonstrated the effectiveness of CIP Loss in producing discriminative shape embeddings.
- Showed compatibility of CIP Loss with existing architectures and loss functions.

## Abstract

Learning discriminative shape representations is a crucial issue for large-scale 3D shape retrieval. In this paper, we propose the Collaborative Inner Product Loss (CIP Loss) to obtain ideal shape embedding that discriminative among different categories and clustered within the same class. Utilizing simple inner product operation, CIP loss explicitly enforces the features of the same class to be clustered in a linear subspace, while inter-class subspaces are constrained to be at least orthogonal. Compared to previous metric loss functions, CIP loss could provide more clear geometric interpretation for the embedding than Euclidean margin, and is easy to implement without normalization operation referring to cosine margin. Moreover, our proposed loss term can combine with other commonly used loss functions and can be easily plugged into existing off-the-shelf architectures. Extensive experiments conducted on the two public 3D object retrieval datasets, ModelNet and ShapeNetCore 55, demonstrate the effectiveness of our proposal, and our method has achieved state-of-the-art results on both datasets.