TTK4900: Hydrogen sulfide detection with behavioural monitoring of salmon juveniles using stereo vision and machine learning

1. Developing dataset of stereo images

• Without Stereo R-CNN: annotator.py
• With Stereo R-CNN: stereorcnn_annotator.py

2. Dataset for Stereo R-CNN

• Training dataset: Stereo_RCNN/data/training_data_stereorcnn
• Testing dataset: Stereo_RCNN/data/testing_data_stereorcnn

3. Stereo camera calibration

• stereo_calibration.py

4. Training Stereo R-CNN

• trainval_net.py
• With Google Colab: setup_train_Stereo_RCNN.ipynb

5. Evaluation of trained Stereo R-CNN model

• test_model.py

6. 3D position estimation of detected objects

• 3D_pos_detections.py

7. Tracking objects with estimation of velocity and acceleration

• Track a specific video over a number of frames: tracker.py
• Track many videos: tracker_all_videos.py

8. Developing dataset of distribution positional data

• sliding_window.py

9. Dataset for hydrogen sulfide classification and estimation

• Training dataset: h2s_estimation/data/training_data
• Testing dataset: h2s_estimation/data/testing_data

10. Classification of hydrogen sulfide

• SVM, decision tree, random forest: classify_h2s_classifiers.py
• AutoML sk-learn: classify_automl_sklearn.ipynb

11. Estimation of hydrogen sulfide

• SVM, decision tree, random forest: estimation_h2s_regressors.py
• AutoML sk-learn: estimation_automl_sklearn.ipynb
• H2O AutoML: estimation_automl_H2O.ipynb