Computer Vision / Deep Learning

Gait Abnormality Detection using CNN

A research-focused CNN classification project for detecting gait abnormalities from ground reaction force based gait data.

Main result95.57%CNN accuracy

cross-validation5-fold

F1-score0.96

baseline improvement18-22%

samples1,200+

Problem

Detecting gait abnormalities manually can be slow and inconsistent across samples, especially when patterns need comparative model evidence.

Solution

Designed and evaluated a multi-layer CNN against traditional ML baselines using training curves, confusion matrices, and 5-fold validation.

Outcome

The CNN reached approximately 95.57% accuracy and outperformed classical ML baselines in the reported comparison.

Workflow

01Gait samples

02CNN blocks

03Feature maps

04Validation

05Classification

CNN architecture, model comparison, validation analysis, research write-up

PythonTensorFlowKerasScikit-learnCNNMatplotlib

Traditional ML baselines included KNN, SVM, and Logistic Regression.
Reported CNN accuracy is approximate because the source states it as around 95.57%.
The page avoids medical deployment claims and presents this as research work.

Project media