NepScript Genesis: Neural Architecture Search for Handwritten Devanagari Digit Synthesis
Computer Vision and Pattern Recognition
Summary
The authors developed NepScript Genesis, a system that automatically finds good designs for GANs to create images of handwritten Devanagari digits. They tested five different search methods and found that one called Adaptive Exploration worked best, making much better images than their starting model, and doing it quickly. They also added special tricks based on the script’s features to avoid common training problems. Finally, when they used the generated images to help train a digit classifier with few real examples, the classifier got noticeably more accurate, showing their approach makes useful fake data.
Authors
Mausam Gurung, Prabin Neupane, Sajjan Acharya
Abstract
This paper introduces NepScript Genesis, a Neural Architecture Search (NAS) framework for automated Generative Adversarial Network (GAN) discovery, applied to conditional Devanagari handwritten digit synthesis. We compare five NAS strategies against a carefully constructed Deep Convolutional GAN (DCGAN) baseline (FID=332.28). Architecture selection utilizes a two-stage pipeline guided by a novel domain-aware evaluation metric (Enhanced Score). Results demonstrate that Adaptive Exploration achieves the optimal quality-efficiency trade-off, attaining an FID of 79.12 -- a 76.19% improvement over the baseline -- and the highest mode coverage among the NAS strategies (Recall=0.531) in under one GPU-hour. Furthermore, we demonstrate that incorporating script-specific structural heuristics into the search phase prevents early-stage mode collapse. In a downstream low-resource evaluation, augmenting 250 real training samples per class with GAN-generated digits from the best NAS model improves CNN classification accuracy from 91.0% to 96.5% (+5.5 percentage points), demonstrating that NAS-optimized synthesis produces digits of sufficient quality to benefit practical recognition pipelines when real data is scarce.