# Rethinking Multi-Branch and Cross-Backbone Fusion for Vehicle Re-Identification under Foundation-Model Pretraining

arXiv: https://arxiv.org/abs/2607.22068 · HTML: https://mlsanity.com/paper/2607.22068/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 27 of its 41 references that are in the map.
- First posted on arXiv 24 Jul 2026.

## Builds on (most cited first, 27 of 27)

- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles](https://mlsanity.com/paper/1612.01474/) (2016)
- [A ConvNet for the 2020s](https://mlsanity.com/paper/2201.03545/) (2022)
- [BEiT: BERT Pre-Training of Image Transformers](https://mlsanity.com/paper/2106.08254/) (2021)
- [Visualizing and Understanding Convolutional Networks](https://mlsanity.com/paper/1311.2901/) (2013)
- [Similarity of Neural Network Representations Revisited](https://mlsanity.com/paper/1905.00414/) (2019)
- [In Defense of the Triplet Loss for Person Re-Identification](https://mlsanity.com/paper/1703.07737/) (2017)
- [Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distribution](https://mlsanity.com/paper/2202.10054/) (2022)
- [Do Vision Transformers See Like Convolutional Neural Networks?](https://mlsanity.com/paper/2108.08810/) (2021)
- [Fine-tuning CNN Image Retrieval with No Human Annotation](https://mlsanity.com/paper/1711.02512/) (2017)
- [Deep Ensembles: A Loss Landscape Perspective](https://mlsanity.com/paper/1912.02757/) (2019)
- [TransReID: Transformer-based Object Re-Identification](https://mlsanity.com/paper/2102.04378/) (2021)
- [Bag of Tricks and A Strong Baseline for Deep Person Re-identification](https://mlsanity.com/paper/1903.07071/) (2019)
- [Circle Loss: A Unified Perspective of Pair Similarity Optimization](https://mlsanity.com/paper/2002.10857/) (2020)
- [ResNet strikes back: An improved training procedure in timm](https://mlsanity.com/paper/2110.00476/) (2021)
- [Learning Discriminative Features with Multiple Granularities for Person Re-Identification](https://mlsanity.com/paper/1804.01438/) (2018)
- [Re-ranking Person Re-identification with k-reciprocal Encoding](https://mlsanity.com/paper/1701.08398/) (2017)
- [Do Wide and Deep Networks Learn the Same Things? Uncovering How Neural Network Representations Vary with Width and Depth](https://mlsanity.com/paper/2010.15327/) (2020)
- [CLIP-ReID: Exploiting Vision-Language Model for Image Re-Identification without Concrete Text Labels](https://mlsanity.com/paper/2211.13977/) (2022)
- [Graph Convolution Based Efficient Re-Ranking for Visual Retrieval](https://mlsanity.com/paper/2306.08792/) (2023)
- [A Comprehensive Survey on Deep-Learning-based Vehicle Re-Identification: Models, Data Sets and Challenges](https://mlsanity.com/paper/2401.10643/) (2024)
- [Unity is Strength: Unifying Convolutional and Transformeral Features for Better Person Re-Identification](https://mlsanity.com/paper/2412.17239/) (2024)
- [Strength in Diversity: Multi-Branch Representation Learning for Vehicle Re-Identification](https://mlsanity.com/paper/2310.01129/) (2023)
- [CLIP-SENet: CLIP-based Semantic Enhancement Network for Vehicle Re-identification](https://mlsanity.com/paper/2502.16815/) (2025)
- [A heterogeneous branch and multi-level classification network for person re-identification](https://mlsanity.com/paper/2006.01367/) (2020)
- [Large-scale Fully-Unsupervised Re-Identification](https://mlsanity.com/paper/2307.14278/) (2023)

## Abstract (arXiv)

Multi-branch architectures and CNN-Transformer fusion are widely believed to improve vehicle re-identification (Re-ID) by combining complementary representations. We revisit this for a DINOv3-pretrained backbone. A single DINOv3-pretrained ConvNeXt with a tuned recipe reaches 88.19 mAP on VeRi-Wild Small and 77.47 on Large from visual cues alone, within the combined evaluation and optimization noise of the strongest protocol-verified metadata-dependent multi-branch baseline, and 92.38/83.68 with training-free re-ranking. Using this baseline and retrieval-level branch diagnostics, we ask whether representational diversity still pays at this scale. In our runs, it does not. Across both benchmarks and every converged configuration, concatenating multiple heads over a shared backbone moves the best single head by under one mAP point in either direction while costing four times the embedding dimension; 99.7% of the concatenation's variance lies in 512 principal components, so the heads not only duplicate one another but each occupies a quarter of its nominal 2048 dimensions. Pushing diversity to its architectural limit, CNN versus Transformer, we grant fusion every advantage through an asymmetric frozen-anchor scheme. Every Transformer configuration still lands at least 13 mAP below the ConvNeXt backbone (13-15 for the two strongest, up to 46 for the weakest), and a paired per-query bootstrap bounds the fusion gain at +0.11 mAP (95% CI) even for the most favourable snapshot we obtained. One strong backbone with the right recipe and re-ranking is the efficiency frontier. All results use single-seed training and one foundation-model family; differences of this size are therefore reported as bounds rather than orderings, and we list falsifiers.