This submission contains data collected during a four-month field demonstration (July-October 2024) of MarineSitu's Adaptable Monitoring Package (AMP) at the Pacific Northwest National Laboratory (PNNL) Marine and Coastal Research Laboratory (MCRL) in Sequim, WA. The project evaluated the platform's survivability in a tidal channel and the iterative development methodology for real-time AI-driven environmental monitoring. The dataset includes: Sample Optical Imagery: High-resolution greyscale images from a modular camera systems capturing confirmed or suspected marine life interactions using automated object detection-activated acquisition. Machine Learning Artifacts: Object detection model weights (.pt) for three iterations each of optical and acoustic models, validation datasets with manually labeled annotations (.txt), and model-generated detection predictions (.json). Analysis Software: Python-based tools for model performance validation (Precision, Recall, mAP50, false positive rate analysis) and synchronized multi-instrument data review. Prerequisites/Assumptions: Use of the included Python scripts requires a Python 3.10+ environment. Users should refer to the provided README.md files within the "Data Viewer" and "Model Validation" directories for installation and usage instructions. Note: This submission includes a representative sample of optical imagery. A larger dataset of optical and sonar data will be made available on July 1, 2029, in the following repository: https://mhkdr.openei.org/submissions/689 (linked below).