CIFAR-10 and EMNIST Model Performance Benchmarks Using VGG-16

Relative performance metrics for deep learning models on two standard computer vision datasets. The data includes test accuracy, inference time, and energy consumption, expressed as percentual differences from a benchmark. Author Amirhossein Douzandeh Zenoozi published the 5.5 KB Excel file on figshare in March 2026.

Use Cases

• Compare the accuracy-efficiency trade-off of different model configurations based on the relative performance metrics.
• Benchmark inference energy consumption for computer vision models based on the total inference energy (TIE) metric.
• Analyze the relationship between inference speed and model sparsity based on the average inference time (AIT) data.
• Identify optimal model configurations for energy-aware deployment based on the highlighted 'best trade-off' rows.

Strengths

• Data is focused on three key performance metrics (accuracy, time, energy) for direct comparison.
• Results are benchmarked against a standard, allowing for relative performance assessment.
• The dataset explicitly highlights the best trade-off and best results for each metric.

Limitations

• The dataset is very small at 5.5 KB, indicating limited scope.
• Row count and specific column definitions are unknown, requiring inspection after download.
• The description lacks detail on the exact benchmark values or the specific model configurations tested.

Provenance

Source

figshare, author Amirhossein Douzandeh Zenoozi.

Collection Method

Likely contains results from experimental evaluations of models on the CIFAR-10 and EMNIST datasets using the VGG-16 architecture.

Freshness

Last updated 2026-03-18 17:44:19.