Skip to content
DraconiticResearch
Search Sources API
182,544 studies31 sourcesUpdated 22 min ago
Peer-reviewedCC BY 4.0PLOS One

Fine-grained Chinese dish recognition on CNFOOD-241: A discriminative learning approach for improving category-level discrimination

  • Xin Xiong1
  • Jinqiang Liao
  • Xiao Mo
  • JianFeng He
  1. 1Faculty of Information Engineering and Automation, Kunming University of Science and Technology, Kunming, China
Open at sourcePDFDOICite

Abstract

Accurate recognition of food images is essential for image-based dietary assessment, nutritional tracking, and digital health applications. However, closed-set Chinese dish recognition remains challenging because many categories in the CNFOOD-241 dataset exhibit strong visual similarity, substantial intra-class variation, and subtle inter-class differences. In this study, we present an integrated discriminative training strategy for fine-grained Chinese dish recognition using RegNetY-32GF as the student backbone. Rather than proposing a new backbone architecture, the study focuses on improving category-level discrimination under a fixed convolutional architecture by combining teacher-guided knowledge distillation, margin-based classification, triplet-based embedding regularization, and mixed-sample augmentation. Across five random seeds, the proposed full model achieved 84.37 + /- 0.29% Top-1 accuracy, 97.66 + /- 0.11% Top-5 accuracy, 83.41 + /- 0.57% macro-F1, and 84.27 + /- 0.30% weighted-F1. Paired bootstrap analysis showed a positive Top-1 improvement of 0.651 percentage points over the baseline, with a 95% confidence interval of [0.491, 0.817], and the mean class-wise Delta F1 across five seeds was significantly greater than zero according to the Wilcoxon signed-rank test (p = 0.017803). These findings suggest that the proposed framework provides a modest improvement in category-level discrimination on CNFOOD-241, with statistically supported evidence at the dataset level, although the magnitude of improvement varies across individual categories. The training code, evaluation scripts, derived result files, and trained model weights are publicly available at the project repository and GitHub Release.

Related

  1. Object recognition in autism: Shape configuration and Category-specificity
    Christian Gerlach, Esben Helby, Linnea Brønnum Hansen +2
    PsyArXivPreprintCC BY 4.0Full textPsychiatry & mental healthOSF
  2. MA-Pose and DMMS-GCN: skeleton-based martial-arts category recognition with edge-device deployment
    Longying Wang, Ruijie Zhang, Huiteng Xu +1
    Frontiers in NeuroroboticsPeer-reviewedCC BYFull textGeneral medicineDOAJ
  3. Rumble Recognition: a gastrointestinal sound discrimination task to measure gastrointestinal interoception
    Hannah S. Savage, Beth F. Longley, Hugo D. Critchley +6
    Scientific ReportsPeer-reviewedCC BYFull textGeneral medicineDOAJ
  4. Integrated proteomics and machine learning reveal AHNAK as a discriminative biomarker of atrial fibrillation in heart failure.
    Wengen Zhu, Zhenbang Gu, Yuwen Huang +5
    BMC medicinePeer-reviewedCC BY-NC-NDFull textCardiovascular medicineEurope PMCDOAJ
  5. Closing the Gap Between Risk and Recognition: Improving Delirium Screening in the Pediatric Intensive Care Unit.
    Anneliese J Grewing, Kelli Jo Welter, Sherrill Caprarola +12
    Pediatric quality & safetyPeer-reviewedCC BY-NC-NDFull textGeneral medicineEurope PMCPMC

Cite

BibTeXRIS
BibTeX
@article{xiong2026finegrained,
  title = {{Fine-grained Chinese dish recognition on CNFOOD-241: A discriminative learning approach for improving category-level discrimination}},
  author = {Xin Xiong and Jinqiang Liao and Xiao Mo and JianFeng He},
  journal = {PLOS One},
  year = {2026},
  month = sep,
  doi = {10.1371/journal.pone.0358248},
  url = {https://doi.org/10.1371/journal.pone.0358248},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - Fine-grained Chinese dish recognition on CNFOOD-241: A discriminative learning approach for improving category-level discrimination
AU  - Xin Xiong
AU  - Jinqiang Liao
AU  - Xiao Mo
AU  - JianFeng He
PY  - 2026
DA  - 2026/09/25
JO  - PLOS One
DO  - 10.1371/journal.pone.0358248
UR  - https://doi.org/10.1371/journal.pone.0358248
AB  - Accurate recognition of food images is essential for image-based dietary assessment, nutritional tracking, and digital health applications. However, closed-set Chinese dish recognition remains challenging because many categories in the CNFOOD-241 dataset exhibit strong visual similarity, substantial intra-class variation, and subtle inter-class differences. In this study, we present an integrated discriminative training strategy for fine-grained Chinese dish recognition using RegNetY-32GF as the student backbone. Rather than proposing a new backbone architecture, the study focuses on improving category-level discrimination under a fixed convolutional architecture by combining teacher-guided knowledge distillation, margin-based classification, triplet-based embedding regularization, and mixed-sample augmentation. Across five random seeds, the proposed full model achieved 84.37 + /- 0.29% Top-1 accuracy, 97.66 + /- 0.11% Top-5 accuracy, 83.41 + /- 0.57% macro-F1, and 84.27 + /- 0.30% weighted-F1. Paired bootstrap analysis showed a positive Top-1 improvement of 0.651 percentage points over the baseline, with a 95% confidence interval of [0.491, 0.817], and the mean class-wise Delta F1 across five seeds was significantly greater than zero according to the Wilcoxon signed-rank test (p = 0.017803). These findings suggest that the proposed framework provides a modest improvement in category-level discrimination on CNFOOD-241, with statistically supported evidence at the dataset level, although the magnitude of improvement varies across individual categories. The training code, evaluation scripts, derived result files, and trained model weights are publicly available at the project repository and GitHub Release.
SN  - 1932-6203
ER  -