2026 04 04 SpeciesNet Results for 2013 07 29 VID00002

This document (now at rev2) summarizes the frame-by-frame analysis of video VID00002, recorded on 2013-07-29, using the MegaDetector object detector and the SpeciesNet classifier. The analysis documents how this AI-based automated detection system responds to the subject visible in the video frames. The analysis includes: 111 full, unmodified frames, 111 wide-crop frames centered on the region where a human observer can visually identify the subject, SpeciesNet JSON outputs, Annotated images with bounding boxes, A structured table of detection outcomes, on-site height verification from a follow-up visit. The data and analysis establish: 1. Full-frame analysis shows detector limitations under motion blur and occlusion, with many noise detections but no detection of the focal subject. 2. Cropped-frame analysis reveals consistent detection of the subject, with alternating animal and human labels. 3. SpeciesNet abstains, indicating the subject does not match any known species in its training taxonomy. 4. The detector’s behavior independently corroborates the human observer’s identification of an object that appears to be an animal. 5. The results directly rebut claims of pareidolia or subjective interpretation. 6. On-site height verification demonstrates that the subject is approximately 8 feet tall and therefore not human.