Benchmarking hybrid CNN and transformer backbones with graph convolution networks (GCN) for flower growth-stage classification
Aritra Das, Karib Shams, Mohammad Rifat Ahmmad Rashid, Raihan Ul Islam, Shamim H. Ripon, Ahmed Wasif Reza · 2026
Read Benchmarking hybrid CNN and transformer backbones with graph convolution networks (GCN) for flower growth-stage classification by Aritra Das, Karib Shams, Mohammad Rifat Ahmmad Rashid, Raihan Ul Islam, Shamim H. Ripon, Ahmed Wasif Reza (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13527139.txt.
Description
Aritra Das and colleagues propose, in a 2026 study, a hybrid image model for classifying flower growth stages. The task is fine-grained, since early, mid, and full bloom differ subtly and labeled data are scarce. Their framework pairs deep image backbones with graph convolutional networks that reason over a grid of image regions.
A new dataset of 3,889 images covers eight Bangladeshi flower species across three stages, forming twenty-four classes. Systematic trials varied graph depth, node resolution, and how neighboring regions connect. Two hybrid backbones reached ninety-seven percent accuracy, and a transformer backbone jumped from eighty-four to ninety-seven percent once graph processing was added. Grid-based graphs beat nearest-neighbor graphs for stability, and moderate graph depth worked best. Cross-checks on a standard Oxford flower dataset supported generalization.
Catalog record
| CREATOR | Aritra Das, Karib Shams, Mohammad Rifat Ahmmad Rashid, Raihan Ul Islam, Shamim H. Ripon, Ahmed Wasif Reza |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 98747 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |