DETECTION
Locates multiple pills in a single image.

INTELLIGENCE
Axclep Intelligence is Axclepmedi's in-house Medication Safety AI engine. It identifies the medications patients bring in from photos and delivers the results as drug safety information.
The goal is not object detection for its own sake, but carrying through to the steps that verifying actual medications requires.

Locates multiple pills in a single image.
Classifies identification candidates for each pill.
Adds drug information to the identification results.
Provides the safety information the review needs.
Supports pharmacist review and clinical judgment.
On the same set of 100 pill types, we compared Top-1 accuracy together with processing time per pill.

Top-1 accuracy (100 pill types)
Processing time per pill
Designed to deliver higher accuracy than general-purpose multimodal models while cutting processing time by more than 100×.
Recognition and analysis run on our in-house AI engine.
Runs on-premise without calling external commercial AI models.
Can be built so sensitive data never leaves the hospital or the country.
We publish performance figures only when the model version, evaluation dataset, split method and metric definitions are ready to publish alongside them.
We built our own image dataset covering more than 10,000 medications at partner pharmacies.
Review context from 10 advisory pharmacists informs the product design.
We never compare figures from different datasets directly, and we publish the evaluation conditions with every result.
Axclep Intelligence is the Medication Safety AI engine developed in-house by Axclepmedi. It locates multiple pills in a photo, classifies identification candidates for each one, and adds drug safety information to support pharmacist review and clinical judgment.
On a set of 100 pill types it reached 98% Top-1 accuracy at 1 second per pill. General-purpose multimodal models compared on the same set scored Gemini 3.1 Pro 91% at 143 seconds, GPT-5.4 90% at 102 seconds, and Opus 4.6 86% at 171 seconds.
Our engine does not call external commercial AI models, so it can run in the hospital's internal (on-premise) environment and can be built so that sensitive data is never sent outside the hospital or abroad.
It is built on our own image dataset of more than 10,000 medications, collected at partner pharmacies, and references public drug information from the Health Insurance Review & Assessment Service, the Ministry of Food and Drug Safety, and the NIH National Library of Medicine. Review context from 10 advisory pharmacists also informs the product design.
Ask us about on-premise configuration, system integration or joint validation.