metadata

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.