Learning Compatible Embeddings

By Qiang Meng, Chixiang Zhang, Xiaoqiang Xu, Feng Zhou. 2021

44 citations recorded by the source corpus.

Achieving backward compatibility when rolling out new models can highly reduce costs or even bypass feature re-encoding of existing gallery images for in-production visual retrieval systems. Previous related works usually leverage losses used in knowledge distillation which can cause performance degradations or not guarantee compatibility. To address these issues, we propose a general framework called Learning Compatible Embeddings (LCE) which is applicable for both cross model compatibility and compatible training in direct/forward/backward manners. Our compatibility is achieved by aligning class centers between models directly or via a transformation, and restricting more compact intra-class distributions for the new model. Experiments are conducted in extensive scenarios such as changes of training dataset, loss functions, network architectures as well as feature dimensions, and demonstrate that LCE efficiently enables model compatibility with marginal sacrifices of accuracies. The code will be available at https://github.com/IrvingMeng/LCE.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Han-Jia Ye
2. De-chuan Zhan
3. Da-Wei Zhou
4. Fu-Yun Wang
5. Liang Ma
6. Lei Yang
7. Jiahuan Zhou
8. Yuxin Peng

Citations in the corpus

• Forward Compatible Few-Shot Class-Incremental Learning (2022, background): “Recent work focuses on improving model’s backward compatibility [7, 28, 35], while we are the first to address the model’s forward compatibility in FSCIL.”
• Learning Continual Compatible Representation for Re-indexing Free Lifelong Person Re-identification (2024, background): “Besides, in the context of data privacy concerns [25], re-indexing is legally forbidden due to the impracticality of storing and re-indexing unauthorized raw images within privacy-sensitive scenarios [1, 4, 14], also known as Re-indexing Free Lifelong ReID (RFL-ReID), as shown in Fig.”
• Metric Compatible Training for Online Backfilling in Large-Scale Retrieval (2023, unknown): “LCE [13] introduces an alignment loss to align the class centers between old and new models and a boundary loss that restricts more compact intra-class distributions for the new model.”
• Forward Compatible Training for Large-Scale Embedding Retrieval Systems (2021, unknown): “Other works [8, 30, 47] attempt to construct a unified representation space on which models are compatible.”
• Boundary-aware Backward-Compatible Representation via Adversarial Learning in Image Retrieval (2023, methodology): “The convergence trend of BCT [18], LCE [14],”