Evaluation Dashboard

Precomputed evaluation and benchmark results for Visual Product Search

Back to Search
{% if evaluation %}

Indexed Images

{{ "{:,}".format(evaluation.num_indexed_images) }}

Evaluation Queries

{{ "{:,}".format(evaluation.num_queries) }}

Embedding Dimension

{{ evaluation.embedding_dimension }}

Self-Match Excluded

{% if evaluation.exclude_self_match %}Yes{% else %}No{% endif %}

Article Type Precision@5

{{ evaluation.main_cards.article_precision_5 }}%

Subcategory Precision@5

{{ evaluation.main_cards.subcategory_precision_5 }}%

Gender + Master Category Precision@5

{{ evaluation.main_cards.gender_master_precision_5 }}%

Image-to-Image Retrieval Evaluation

Model: {{ evaluation.model }} | Dataset: {{ evaluation.dataset }}

{% for row in evaluation.rows %} {% endfor %}
Match Type Definition Precision@5 NDCG@10 mAP@10 MRR@10
{{ row.name }} {{ row.definition }} {{ row.precision_5 }}% {{ row.ndcg_10 }}% {{ row.map_10 }}% {{ row.mrr_10 }}%
{% else %}

Evaluation Results Not Available

Run evaluation locally and save the generated JSON file before showing this dashboard. 

python -m visual_product_search.evaluation.evaluate_image_to_image
{% endif %} {% if benchmark %}

Search Benchmark

Search latency measured using precomputed embeddings.

Mean Latency

{{ benchmark.mean }} ms

P50 Latency

{{ benchmark.p50 }} ms

P95 Latency

{{ benchmark.p95 }} ms

P99 Latency

{{ benchmark.p99 }} ms

Top-K {{ benchmark.top_k }}
Measured Runs {{ benchmark.num_runs }}
Embedding File Size {{ benchmark.embedding_file_mb }} MB
Metadata File Size {{ benchmark.metadata_file_mb }} MB
{% else %}

Benchmark Results Not Available

Run benchmark locally after generating embeddings. 

python -m visual_product_search.evaluation.benchmark
{% endif %}