OctMNIST Model Performance: VGG-16 Accuracy, Speed, and Energy Trade-offs

Model performances for the OctMNIST dataset, evaluated using the VGG-16 architecture. The dataset contains test accuracy, average inference time, and total inference energy consumption, expressed as relative percentual differences to benchmark values. It was authored by Amirhossein Douzanneh Zenoozi and last updated on March 18, 2026.

Use Cases

• Benchmarking model efficiency based on relative performance metrics for accuracy, time, and energy.
• Identifying optimal model configurations for a best trade-off between multiple performance metrics.
• Analyzing the impact of pruning techniques like 'Wise Pruning' on inference energy usage and CO2 emissions.

Strengths

• Data is licensed under CC-BY-4.0, allowing for open sharing and modification.
• Performance metrics are expressed as relative differences, facilitating direct comparison to a benchmark.

Limitations

• The dataset is very small at 5.5 KB, indicating limited scope.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

figshare, authored by Amirhossein Douzanneh Zenoozi.

Collection Method

Likely contains results from experimental evaluations of the VGG-16 architecture on the OctMNIST dataset.

Freshness

Last updated 2026-03-18 17:44:20; freshness should be verified.