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