---
tags:
- image-classification
- multi-task-learning
- keras
- medical
- diagnostics
- drug-testing
- alcohol-testing
library_name: keras
datasets:
- custom-dataset
license: apache-2.0
model_name: DrugTest_AI
---

# Drug and Alcohol Test Classification Model

The model analyzes images of test strips, classifies the type of test (drug or alcohol), and provides a corresponding result (e.g., Positive/Negative/Invalid or BAC level). This can be used in medical diagnostics, workplace drug testing, or other contexts where rapid test result analysis is required.

## Key Features of the Model:
- **Multi-Task Learning**: The model performs two classification tasks:
  1. Predicting the Drug Type.
  2. Predicting the Test Result (including BAC levels for alcohol).
- **Architecture**: It uses a shared backbone (InceptionResNetV2 pretrained on ImageNet) for feature extraction, followed by two separate dense layers for each task.
- **Custom Data Generators**: These split the labels into two parts (Drug Type and Test Result) and one-hot encode them for multi-class classification.
- **Input Data**: The model processes images of test strips, which are resized to (224, 224, 3) for consistency.

## Drug Test Classification:
- The model classifies the type of drug being tested (e.g., AMP, BAR, BUP, COC, etc.) based on test strip images.
- It also determines the result of the test for each drug (Positive, Negative, or Invalid).

## Alcohol Test Classification:
- For alcohol tests, the model uses the Blood Alcohol Concentration (BAC) levels, which are treated as distinct classes.

## Note:
- The model is still in Beta phase.