Visual Finder｜Object Detector vs Flagfall

Visual Finder detects people and objects in your iPhone or iPad camera feed in real time, letting you review, track, and record results as they happen. It supports 80 COCO-based categories, including people, vehicles, animals, and everyday items. Switch between Box, which marks objects with bounding boxes, and Mask, which highlights each object’s shape. Track detected objects, revisit them as cropped images, review each session through statistics and charts, and export results as CSV. Detection runs with Core ML directly on your device. —————————— ■ Real-time object detection Detect objects live through your camera. Choose only the categories you need or select all available categories. Each result shows its location, category, and confidence. The screen also displays the current detection count and inference FPS. —————————— ■ Box and Mask display modes Box surrounds each object with a bounding box for a clear view of its location and size. Mask uses instance segmentation to show the region along the object’s shape, revealing contours and occupied areas that a rectangle alone cannot capture. Use both modes to compare the same scene from different perspectives. —————————— ■ Object tracking Follow moving objects across camera frames. Each tracked object receives an ID, making it easy to distinguish multiple objects in the same category. —————————— ■ Cropped detection images Recent detections appear as object crops at the bottom of the screen. Tap a crop for a larger view and save selected images to your photo library. In Mask mode, crops follow the segmented shape. They can be saved as PNG files with everything outside the object made transparent. You can also adjust the history limit and haptic feedback for new crops. —————————— ■ Session summaries Results from start to stop are grouped into one session. Logs include: • Detection duration • Processed frames and total detections • Detected categories and counts by category • Detected frames and appearance rate • Average confidence and detection area • Maximum detections per frame • Average FPS Charts show changes over time and each category’s share. Review not only what appeared, but how frequently it was detected. —————————— ■ CSV export Export session logs in two formats: Summary, with overall and category statistics, and Data, with results for each frame. Share files through the iOS or iPadOS share sheet for use with spreadsheets and analysis tools. —————————— ■ Adjustable detection settings Fine-tune the maximum detection rate, confidence threshold, IoU threshold, minimum detection size, and Object Tracking. Choose 1, 5, 10, 20 FPS, or Max. Actual inference speed depends on the device, model, scene, and device conditions. —————————— ■ Camera selection Use available Wide, Ultra Wide, Telephoto, and Front cameras. Compatible external cameras exposed through AVCapture can also be selected. Camera availability and behavior vary by device, OS, and connected camera. —————————— ■ On-device processing Visual Finder uses Core ML. Camera frames are processed on your device and do not need to be sent to an external AI server for object detection. —————————— ■ Ideal for Explore AI, object detection, instance segmentation, and tracking with a live camera. Compare Box and Mask results, save object crops, analyze sessions or CSV data, and use the app for research, development, learning, demos, or external-camera tests. Some detection categories, Object Tracking, and ad removal require an in-app purchase. Results vary with device performance, lighting, object size and orientation, motion, occlusion, and other conditions. This app is not intended for medical, safety management, identity verification, or other uses where false or missed detections could have serious consequences. Terms of Use (Apple Standard EULA): https://www.apple.com/legal/internet-services/itunes/dev/stdeula/

Flagfall is a chess clock built for real games — at the club, at a tournament, or across the kitchen table. TIME CONTROLS • Fischer increment — time added after every move • Bronstein delay — time spent is given back, up to the increment • US delay — the clock only starts counting once the delay has elapsed • Byo-yomi — main time followed by periods, for go and shogi • Multi-stage FIDE controls — 40 moves in 90 minutes, then 30 minutes, with a 30-second increment from move one • Asymmetric time — give one player a handicap BUILT FOR THE TABLE The whole half of the screen is the tap target, not a small button. In table mode the top panel rotates 180 degrees for the player sitting opposite, and each side sees the opponent's remaining time the right way up. The screen stays awake for the entire game. Your settings and the game in progress survive closing the app — reopen it and the clock is exactly where you left it, paused. ACCURATE BY DESIGN The clock never decrements a counter on a timer. Remaining time is derived from timestamps, so a backgrounded app, a late frame or a locked screen cannot make it drift. NO ADS, NO ACCOUNT, NO DATA Flagfall makes no network requests. It works entirely offline, collects nothing, and contains no advertising or tracking whatsoever. There is no account to create and nothing to subscribe to. Available in English, French, Spanish, German, Italian, Portuguese, Dutch, Polish, Russian, Turkish, Japanese and Chinese.