652. Edge-Optimized Computer Vision for Financial Inclusion: Dendritic Neural Networks in Infrastructure-Constrained Retail Environments
Invited abstract in session Business Management in Dynamic Emerging Markets, stream Selected Aspects of International Finance and OR.
Authors (first author is the speaker)
| 1. | Helper Zhou
|
| School of Accounting, Economics and Finance, University of KwaZulu Natal | |
| 2. | Alford Toruvanda
|
| Research and Development, QUANTLYTIX | |
| 3. | Gordon Dash
|
| Finance and Decision Sciences, University of Rhode Island | |
| 4. | Nina Kajiji
|
| Computer Science and Statistics, University of Rhode Island, and The NKD Group, Inc. |
Abstract
Informal retail sustains billions in emerging markets, but digital POS systems often fail due to unstable connectivity, low literacy, and dynamic inventories. This paper introduces a camera-first, offline POS using smartphones for inventory tracking without barcodes or SKUs. Using vegetable recognition, we evaluate dendritic optimization via PerforatedAI to enhance lightweight vision models under edge constraints. MobileNetV3-Small achieved 97.44% accuracy (vs. 95.16%), a 47.1% error reduction without added complexity. Results show impact of dendritic optimization.
Keywords
- Artificial Intelligence
- Big Data
- Optimization Models and Methods
Status: accepted
Back to the list of papers