Label 10k Dart Tip Images
Budget: £20 – £250 GBP
Dart Tip Labeling — Click-to-Mark Precision Task
I have a purpose-built web-based labeling tool that will be loaded with approximately 10,000 cropped dartboard images from multiple camera angles across several dart scoring rigs. Each image shows a section of a dartboard with one or more darts visible.
Your task: For each image, click the exact pixel where the dart tip contacts the board surface. One click per image — that's it. The tool records the coordinates automatically and advances to the next image.
There are three possible actions per image:
Click the tip (~85% of images) — Click precisely where the metal tip meets the board surface.
"Tip Not Visible" (~12%) — A dart is present but the tip is obscured, out of frame, or too blurry to pinpoint. One button press.
"No Dart" (~3%) — The image doesn't contain a visible dart. One button press.
What you'll see: Cropped images of a dartboard from various camera angles (side-on, below, angled). The dart barrel and flights are clearly visible. You need to identify and click the very end of the metal tip — the point closest to the board surface. Precision matters: we need accuracy within 2-3 pixels.
The tool provides:
A "Before" panel showing the board without the new dart, for reference
A "Diff" panel showing only the new dart isolated from any previous darts already on the board
Filter controls to work through batches by priority (multi-dart scenes, specific rigs, low-confidence detections)
Undo, skip, and keyboard shortcuts for speed (click = label tip, D = no dart, V = not visible, Z = undo, arrow keys = navigate)
Auto-advance after each label
Important detail: Some images show 2-3 darts in the frame from the same turn. You must label only the newest dart's tip. The "Diff" panel on the right highlights which dart is new (it appears as a bright silhouette on a black background). If there's no Diff panel, there's only one dart to label.
Access: Secure web URL with password login. Works in any modern browser, no software install required. You can work from anywhere at your own pace.
Deliverable:
Completed pass through all ~10,000 images with accurate tip position labels
Expected throughput: 2-4 seconds per image once familiar with the tool (experienced users average under 2 seconds)
Full batch completable in approximately 8-12 hours of focused work
Acceptance criteria:
Tip placement accuracy within 3 pixels of the actual board contact point, validated against a gold-standard calibration set
Agreement with calibration batch ≥ 95%
No unlabeled images remaining in the queue
What this is for: These labels train an AI model to automatically detect dart tip positions for an electronic dart scoring system. Your precision directly impacts scoring accuracy.
Ongoing opportunity: New images arrive daily (100-500 per day) from live dart scoring systems across multiple venues. Consistent, reliable labelers will be offered ongoing batches.
I have a purpose-built web-based labeling tool that will be loaded with approximately 10,000 cropped dartboard images from multiple camera angles across several dart scoring rigs. Each image shows a section of a dartboard with one or more darts visible.
Your task: For each image, click the exact pixel where the dart tip contacts the board surface. One click per image — that's it. The tool records the coordinates automatically and advances to the next image.
There are three possible actions per image:
Click the tip (~85% of images) — Click precisely where the metal tip meets the board surface.
"Tip Not Visible" (~12%) — A dart is present but the tip is obscured, out of frame, or too blurry to pinpoint. One button press.
"No Dart" (~3%) — The image doesn't contain a visible dart. One button press.
What you'll see: Cropped images of a dartboard from various camera angles (side-on, below, angled). The dart barrel and flights are clearly visible. You need to identify and click the very end of the metal tip — the point closest to the board surface. Precision matters: we need accuracy within 2-3 pixels.
The tool provides:
A "Before" panel showing the board without the new dart, for reference
A "Diff" panel showing only the new dart isolated from any previous darts already on the board
Filter controls to work through batches by priority (multi-dart scenes, specific rigs, low-confidence detections)
Undo, skip, and keyboard shortcuts for speed (click = label tip, D = no dart, V = not visible, Z = undo, arrow keys = navigate)
Auto-advance after each label
Important detail: Some images show 2-3 darts in the frame from the same turn. You must label only the newest dart's tip. The "Diff" panel on the right highlights which dart is new (it appears as a bright silhouette on a black background). If there's no Diff panel, there's only one dart to label.
Access: Secure web URL with password login. Works in any modern browser, no software install required. You can work from anywhere at your own pace.
Deliverable:
Completed pass through all ~10,000 images with accurate tip position labels
Expected throughput: 2-4 seconds per image once familiar with the tool (experienced users average under 2 seconds)
Full batch completable in approximately 8-12 hours of focused work
Acceptance criteria:
Tip placement accuracy within 3 pixels of the actual board contact point, validated against a gold-standard calibration set
Agreement with calibration batch ≥ 95%
No unlabeled images remaining in the queue
What this is for: These labels train an AI model to automatically detect dart tip positions for an electronic dart scoring system. Your precision directly impacts scoring accuracy.
Ongoing opportunity: New images arrive daily (100-500 per day) from live dart scoring systems across multiple venues. Consistent, reliable labelers will be offered ongoing batches.
Related categories:
Data Entry
Image Processing
Computer Vision
Data Collection
Data Management
Image Analysis