Figure 7
From: FathomNet: A global image database for enabling artificial intelligence in the ocean
Comparison of activity recognition output from object detection (a) and a human expert (c) for one video collected by NOAA’s ROV Deep Discoverer during the CAPSTONE expedition45. In one instance (b), the algorithm (indicated by a bounding box) correctly identifies an animal with low contrast (image has been contrast enhanced). In another (d), the algorithm fails to identify an animal that is blurry due to unfocused optics.
