**Intro** -
CIFAR-10 is a standard dataset used to benchmark image classification models. This hackathon submission tests adding Perforated AI dendrites (Dendritic Optimization) to a ResNet18-based CIFAR-10 classifier in PyTorch, and compares its performance against the same baseline ResNet18 model trained without dendrites.
Team:
Yatharth Khanna - Student/Founder@Meta Infinium (https://www.linkedin.com/in/yatharth-khanna-960ab224b/)
**Project Impact** -
Improving the accuracy of an image classification model matters because it reduces incorrect predictions and improves reliability in real-world applications like robotics, smart cameras, edge-device vision, and automated quality inspection. Even small improvements in test accuracy can meaningfully reduce remaining error and increase confidence in deployment scenarios where consistent recognition performance is important.
**Usage Instructions** -
Installation:
pip install -r requirements.txt
Run Baseline (No Dendrites):
python cifar10_resnet18_baseline.py
Run Dendritic Optimization (PerforatedAI):
python cifar10_resnet18_perforatedai.py
**Results**
This project compares the final test accuracy of the ResNet18 CIFAR-10 baseline model vs the same architecture trained with PerforatedAI dendrites.
Model Final Test Accuracy Notes
Traditional (No Dendrites) 85.71% Baseline ResNet18 CIFAR-10
Dendritic (PerforatedAI) 93.48% ResNet18 + Dendritic Optimization
Remaining Error Reduction:
Baseline Error = 100 - 85.71 = 14.29
Dendritic Error = 100 - 93.48 = 6.52
Remaining Error Reduction = (14.29 - 6.52) / 14.29 × 100 = 54.37%
This provides a Remaining Error Reduction of 54.37%
**Raw Results Graph**
