SAAIL Lab SAAIL Lab
DSA 2026 · Oral Oral Presentation , Data Science Africa 2026, Kampala 1+ Citations Published · Oral Presentation

When Synthetic Data Hurts: DCGAN Augmentation Failure Under Extreme Data Scarcity for Cassava Disease Classification in East Africa

Patrick Vincent Ndowo 1 , Innocent Nyalala 1,2 , Khamis Haji 1
1School of Engineering and Science, Indian Institute of Technology Madras Zanzibar, Tanzania
2Wadhwani School of Data Science and AI (WSDSAI), Indian Institute of Technology Madras, Chennai, India
Data Science Africa (DSA) 2026 Workshop, Makerere University, Kampala, Uganda (Jul 2026)
When Synthetic Data Hurts: DCGAN Augmentation Failure Under Extreme Data Scarcity for Cassava Disease Classification in East Africa

Summary: Macro-F1 degradation caused by DCGAN synthetic augmentation across extreme few-shot regimes (n = 25, 50, 100 real images/class): pretrained ResNet-50 drops by 35.9 pp and EfficientNet-B3 by 32.7 pp.

Abstract

We test whether DCGAN augmentation alleviates labeled-data scarcity for cassava disease classification in Uganda. Training a DCGAN on 25, 50, or 100 real images per class and augmenting four classifier architectures to 500 images per class, we find that GAN augmentation consistently degrades macro-F1: by 23.2 pp at n = 25, 21.2 pp at n = 50, and 10.3 pp at n = 100. Pretrained classifiers suffer far more (ResNet-50: -35.9 pp; EfficientNet-B3: -32.7 pp) than a from-scratch CNN (-1.9 pp). GAN diagnostics confirm discriminator dominance and mode collapse, producing semantically empty images that poison fine-tuning. Within the unregularised DCGAN framework studied, these results establish an empirical minimum viability threshold of roughly 100-200 images per class and counsel caution against synthetic GAN augmentation in extreme few-shot African agricultural datasets.

Keywords: DCGANsynthetic datadata scarcitycassava diseaseEast AfricaDSA 2026

Key Findings & Contributions

-23.2 pp
Mean Drop (n=25)
Synthetic GAN poisons training
-35.9 pp
ResNet-50 Drop
Pretrained hierarchy damage
45 Runs
Controlled Matrix
3 scales x 4 backbones
100-200
Min Real Threshold
Required images per class
  1. Conducted 45 controlled DCGAN augmentation runs across three low-shot regimes (n = 25, 50, 100 real images/class) and four classifiers on Ugandan field cassava pathology.
  2. Demonstrated that unregularized DCGAN augmentation reduces macro-F1 by an average of 23.2 pp at n = 25, 21.2 pp at n = 50, and 10.3 pp at n = 100 compared with geometric augmentation alone.
  3. Identified asymmetric harm to ImageNet-pretrained backbones (-35.9 pp ResNet-50, -32.7 pp EfficientNet-B3) due to discriminator dominance and semantic mode collapse (lossD near 0.000, lossG spiking to 44.5).

Key Empirical Figure

Classifier performance comparing real-only geometric augmentation against real + DCGAN augmentation across n = 25, 50, and 100 real training images per class on iCassava.
Classifier performance comparing real-only geometric augmentation against real + DCGAN augmentation across n = 25, 50, and 100 real training images per class on iCassava.

Benchmark Results

Macro-F1 Delta (pp) from DCGAN Augmentation vs. Geometric-Only Baseline (3 Seeds)

Classifiern = 25 imgs/clsn = 50 imgs/clsn = 100 imgs/cls
Lightweight CNN (scratch)-1.9 ± 2.2-4.5 ± 2.4+0.6 ± 2.8
MobileNetV3-22.1 ± 9.2-20.9 ± 2.2-10.1 ± 8.0
EfficientNet-B3-32.7 ± 8.1-22.0 ± 6.0+0.9 ± 0.7
ResNet-50-35.9 ± 11.8-37.3 ± 4.0-32.5 ± 9.2
4-Model Mean-23.2 ± 15.4-21.2 ± 13.4-10.3 ± 15.7

BibTeX

@inproceedings{ndowo2026dcgan,
  title={When Synthetic Data Hurts: DCGAN Augmentation Failure Under Extreme Data Scarcity for Cassava Disease Classification in East Africa},
  author={Ndowo, Patrick and Nyalala, Innocent and Haji, Khamis},
  booktitle={Data Science Africa (DSA) 2026 Workshop},
  address={Makerere University, Kampala, Uganda},
  year={2026},
  month={July},
  note={Oral Presentation}
}
APA Citation

Ndowo, P., Nyalala, I., & Haji, K. (2026, July). When Synthetic Data Hurts: DCGAN Augmentation Failure Under Extreme Data Scarcity for Cassava Disease Classification in East Africa. Data Science Africa (DSA) 2026 Workshop, Makerere University, Kampala, Uganda. (Oral Presentation)