MiniImageNet Model Benchmark: Inference Time and Memory Usage for 5-Way-1-Shot Tasks

Jianxiong Chen published a benchmark dataset on figshare in April 2026. It records inference time and memory usage for various models evaluated on the MiniImageNet dataset under a 5-way-1-shot learning protocol. The dataset is stored in a 5.5 KB XLS file.

Use Cases

• Compare computational efficiency across different model architectures based on the reported inference times.
• Profile memory requirements for few-shot learning models based on the recorded memory usage metrics.
• Select models for deployment in resource-constrained environments based on the performance trade-offs.
• Validate model optimization techniques by comparing against the provided benchmark results.

Strengths

• Dataset is small (5.5 KB), facilitating quick download and inspection.
• Provides a specific, comparable benchmark for models on a standardized 5-way-1-shot task.
• Released under a permissive CC-BY-4.0 license, allowing for broad reuse.

Limitations

• Row count and specific column definitions are unknown, limiting detailed assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

figshare, author Jianxiong Chen.

Collection Method

Likely contains measured performance metrics from model evaluations.

Time Range

Benchmark data likely corresponds to the author's evaluation period, not explicitly stated.

Freshness

Last updated 2026-04-28 17:43:01; freshness should be verified.

Geography

null