evedaai/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.