Custom Observer App vs Visual Finder

The Custom Observer app, enables data collection using standard methods and protocols to produce credible and actionable insights that address needs and priorities of communities for environmental and ecological monitoring. Our mission is to empower communities to lead in environmental monitoring and research by branding their own app environment in which they can combine diverse data collection methods to drive effective ecosystem stewardship and resource management. The Custom Observer App is a modern approach to caring for and stewarding the land. Data collection priorities are set by communities and collaborating researchers and can range from environmental data related to extreme weather events such as wind speed or storm surge, or ecological indicators such as species distribution and phenology, presence of invasive species or declines in important subsistence species. The app supports the employment of rigorous protocols for data validity and reliability, blending scientific and social methodologies. Moreover, the app supports the assurance of data accuracy by providing detailed quality assurance and control mechanisms. The Custom Observer app is the technical infrastructure for community networks to lead in defining research priorities, generating and managing access to data equitably, and ensuring that future ecological and environmental monitoring strategies, all while also enabling community networks to apply their own branding and look and feel to the app. The Custom Observer App’s core function is to support the implementation of community-driven monitoring programs. Data Collection Programs are sets of related observation activities that group various data collection forms/surveys. The Programs module in the app organizes data collection forms, users, and observations into manageable categories for a defined network of communities and organizations. On the app, an ‘observation’ is a single data point collected in a survey/form related to certain species, environmental variables or events at a specific point in space and time within a data collection program. The mobile app supports the entry of observations while offline, making the system ideal for use in remote areas and tribal communities. The data from observations go through quality assurance and are used in analysis and reporting. Geospatial and mapping functionality supports program effectiveness by allowing users to use relevant GIS map layers and maintain a spatially organized context to their data, contributing to more comprehensive record keeping and informed decision-making.

Visual Finder is an iPhone / iPad app for real-time object detection using the device camera. It lets you find objects in the camera preview, display detection results as bounding boxes or masks, review detected classes, positions, counts, confidence values, cropped object images, and session statistics directly on the device. Main Features * Real-time object detection using the device camera * Bounding box detection mode for fast object localization * Mask detection mode for object region visualization * Support for 81 object classes, including people, vehicles, animals, daily items, food, furniture, electronics, and more * Selection of specific detection targets * Search and filtering of detectable object classes * Live overlay of detected objects on the camera preview * Display of inference FPS, detection count, confidence, and detection status * Adjustable detection speed, confidence threshold, IoU threshold, and minimum object size * Camera switching between available device cameras * Support for compatible external cameras connected to iPad via USB-C * Cropped image preview of detected objects * Saving selected cropped detection images to the photo library * Session logging for detection results * Class-level detection statistics and charts * CSV export of session logs * Settings for logging, crop history, haptic feedback, and display language Visual Finder is designed for checking object detection behavior in real-world scenes, comparing bounding box and mask detection, adjusting detection parameters, and reviewing results after each session. During detection, the app shows where objects are found in the camera image and summarizes what classes appeared, how often they were detected, and how stable the results were during the session. Detected objects can be reviewed as cropped images. In mask mode, object crops can be generated along the detected object region, making it easier to inspect individual detection results or save visual samples for later review. Session logs include information such as the detection mode, selected detection targets, session start and end time, total detections, detected classes, class-level counts, detected frame count, appearance rate, average confidence, average detection area, and maximum detections per frame. Logs can be reviewed inside the app with charts and exported as CSV files for spreadsheet software, analysis tools, or development workflows. The app performs detection locally on the device. Camera frames, detection results, cropped images, and session logs are not uploaded to an external server by the app. Supported Devices * iPhone and iPad models with a supported built-in camera * iPad models that support compatible external cameras connected via USB-C * Devices capable of running the included Core ML object detection models Camera availability, external camera support, detection speed, supported modes, and performance vary depending on the device, connected camera, camera configuration, OS version, and model settings. Detection results are provided for reference, testing, and analysis purposes only. Results may vary depending on device performance, lighting, object size, camera angle, motion, occlusion, and other scene conditions. The app is not intended for medical, safety-critical, legal, surveillance, or identity verification use.