ObjMark: image annotation

[e-mail]

About this app

ObjMark - Image Annotation for Object Detection
Create labeled training datasets for machine learning directly on your phone. ObjMark is a mobile annotation tool built for practitioners who need to label images for object detection models, with direct export support for Google Vertex AI.

Core Features
Bounding Box Annotation Draw precise bounding boxes with intuitive touch controls. Move and resize boxes smoothly without lag. Each box can have multiple labels assigned, with visual indicators showing all active classifications.

Label Management Set up single-class projects (like vehicle detection) or multi-class workflows (where each object has multiple attributes like type, color, and condition). Import existing label sets from CSV files or build your own. Organize labels into categories for complex classification tasks.

Image Sources Add images by taking photos with your camera, importing from your photo library, or browsing device albums. Work with batches of images that need the same type of annotation.

Export Formats Generate JSONL files formatted for Vertex AI object detection, or export to CSV for other platforms. All coordinates are normalized to 0-1 range, making them resolution-independent and ready for training. Preview export data before saving to verify accuracy.

Workflow Tools Navigate between images using left/right arrows. Undo mistakes with the history function. Save label presets to quickly switch between different annotation scenarios. Mark images as complete to track progress through large datasets.

Local Storage Everything stays on your device. No cloud syncing, no accounts, no data collection. Your images and annotations remain private. Export files only when you choose to share them.

Technical Details
Export: JSONL (Vertex AI format), CSV
Coordinates: Normalized (0-1) for any image resolution
Multi-label: Supported per bounding box
Storage: Local only, no internet required
Interface: Light/dark themes
Typical Workflows
Computer Vision Research Label datasets for academic papers or thesis projects. Annotate objects in various domains from medical imaging to satellite imagery.

Custom Model Training Build training data for YOLOv8, Faster R-CNN, or other detection architectures. Compatible with TensorFlow, PyTorch, and cloud ML platforms.

Vertex AI Projects Export directly to Vertex AI format. Upload images to Google Cloud Storage, import the annotation file, and start training without format conversion.

Commercial Applications Create datasets for retail product detection, manufacturing quality control, inventory management, or security system training.

Setup Process
Create a dataset and define your labels
Add images from camera or library
Draw bounding boxes and assign labels
Export annotations in your preferred format
Upload to your training platform
Vertex AI Integration
ObjMark generates annotation files matching Google Vertex AI's object detection specification. After export, upload your images to Google Cloud Storage, update the GCS paths in the JSONL file, and import to Vertex AI for training.

Privacy
All data processing happens on your device. No external servers, no telemetry, no tracking. Uninstall the app to remove all data, or manage datasets individually within the app.

Note: This app creates annotation files. You'll need to upload images and annotations to Google Cloud Storage separately if using Vertex AI. The app doesn't handle cloud uploads directly.

What’s new

The app has been updated to prepare for production release.