Connect a four-class image classifier to reproducible data, experiment and deployment workflows.
02
System architecture
Follow the flow from input to evaluated output. Select a stage to inspect its role.
Image — Image
FastAPIDockerGitHub ActionsAWS ECRAWS EC2
03
Engineering decisions
Engineering decision
Treat deployment as part of ML
Problem
Move beyond notebook-only training while retaining reproducible experiments.
Approach
VGG16 transfer learning, DVC versioning and automated Docker / ECR / EC2 deployment.
Trade-off / scope
Explainability, larger datasets and production monitoring remain future improvements.
04
Evaluation, in context.
direct evidenceModel evaluation
VGG16 transfer learning with targeted fine-tuning; 94% validation accuracy across four categories.
Documented result94% validation accuracyFour-class model validation (local research; completed): Validation accuracy; split sizes and class labels are not supplied. Four poultry disease categories; dataset identity and sample count not supplied. Validation accuracy: 94%. Four-class validation accuracy; sample count and external validation not supplied. Validation accuracy across four classes; dataset size, split details, class names and external validation are not supplied. Sub-500 ms inference statement has no supplied hardware, percentile, sample size or production SLA definition.
supporting evidenceInference statement
Profile states optimization for sub-500 ms inference; measurement definition not supplied.
Documented resultsub-500 ms inference · profile statementProfile-reported inference latency (unspecified; completed): Measurement method, aggregation, hardware and sample count not supplied. Inference workload not specified in the profile. Reported inference latency: sub-500 ms. Profile-reported bound; no p95/p99, hardware, sample count or SLA supplied. Validation accuracy across four classes; dataset size, split details, class names and external validation are not supplied. Sub-500 ms inference statement has no supplied hardware, percentile, sample size or production SLA definition.
Four-class model validation: completed. Four poultry disease categories; dataset identity and sample count not supplied. Profile-reported inference latency: completed. Inference workload not specified in the profile.
05
Reliability & deployment
A documented safety and scope boundary.
Validation accuracy across four classes; dataset size, split details, class names and external validation are not supplied. Sub-500 ms inference statement has no supplied hardware, percentile, sample size or production SLA definition.