|
|
| ---
|
| 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.
|
|
|
|
|
