# TCFormer: A 5M-Parameter Transformer with Density-Guided Aggregation for Weakly-Supervised Crowd Counting

arXiv: https://arxiv.org/abs/2512.22203 · HTML: https://mlsanity.com/paper/2512.22203/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 15 of its 44 references that are in the map.
- First posted on arXiv 21 Dec 2025.

## Builds on (most cited first, 15 of 15)

- [CSRNet: Dilated Convolutional Neural Networks for Understanding the Highly Congested Scenes](https://mlsanity.com/paper/1802.10062/) (2018)
- [TinyViT: Fast Pretraining Distillation for Small Vision Transformers](https://mlsanity.com/paper/2207.10666/) (2022)
- [Composition Loss for Counting, Density Map Estimation and Localization in Dense Crowds](https://mlsanity.com/paper/1808.01050/) (2018)
- [NWPU-Crowd: A Large-Scale Benchmark for Crowd Counting and Localization](https://mlsanity.com/paper/2001.03360/) (2020)
- [Switching Convolutional Neural Network for Crowd Counting](https://mlsanity.com/paper/1708.00199/) (2017)
- [Boosting Crowd Counting via Multifaceted Attention](https://mlsanity.com/paper/2203.02636/) (2022)
- [TransCrowd: weakly-supervised crowd counting with transformers](https://mlsanity.com/paper/2104.09116/) (2021)
- [Towards Using Count-level Weak Supervision for Crowd Counting](https://mlsanity.com/paper/2003.00164/) (2020)
- [Pixel-wise Crowd Understanding via Synthetic Data](https://mlsanity.com/paper/2007.16032/) (2020)
- [Improving Point-based Crowd Counting and Localization Based on Auxiliary Point Guidance](https://mlsanity.com/paper/2405.10589/) (2024)
- [Semi-supervised Crowd Counting via Density Agency](https://mlsanity.com/paper/2209.02955/) (2022)
- [Towards Unsupervised Crowd Counting via Regression-Detection Bi-knowledge Transfer](https://mlsanity.com/paper/2008.05383/) (2020)
- [CrowdMLP: Weakly-Supervised Crowd Counting via Multi-Granularity MLP](https://mlsanity.com/paper/2203.08219/) (2022)
- [Semi-supervised Counting via Pixel-by-pixel Density Distribution Modelling](https://mlsanity.com/paper/2402.15297/) (2024)
- [Robust Zero-Shot Crowd Counting and Localization With Adaptive Resolution SAM](https://mlsanity.com/paper/2402.17514/) (2024)

## Built on by (most cited first, 1 of 1)

- [FLASH: Efficient Impact Fall Detection with Unified Hypergraph State-Space Model](https://mlsanity.com/paper/2607.25791/) (2026)

## Abstract (arXiv)

Crowd counting typically relies on labor-intensive point-level annotations and computationally intensive backbones, restricting its scalability and deployment in resource-constrained environments. To address these challenges, this paper proposes the TCFormer, a tiny, ultra-lightweight, weakly-supervised transformer-based crowd counting framework with only 5 million parameters that achieves competitive performance. Firstly, a powerful yet efficient vision transformer is adopted as the feature extractor, the global context-aware capabilities of which provides semantic meaningful crowd features with a minimal memory footprint. Secondly, to compensate for the lack of spatial supervision, we design a feature aggregation mechanism termed the Learnable Density-Weighted Averaging module. This module dynamically re-weights local tokens according to predicted density scores, enabling the network to adaptively modulate regional features based on their specific density characteristics without the need for additional annotations. Furthermore, this paper introduces a density-level classification loss, which discretizes crowd density into distinct grades, thereby regularizing the training process and enhancing the model's classification power across varying levels of crowd density. Therefore, although TCformer is trained under a weakly-supervised paradigm utilizing only image-level global counts, the joint optimization of count and density-level losses enables the framework to achieve high estimation accuracy. Extensive experiments on four benchmarks including ShanghaiTech A/B, UCF-QNRF, and NWPU datasets demonstrate that our approach strikes a superior trade-off between parameter efficiency and counting accuracy and can be a good solution for crowd counting tasks in edge devices.