Show-Jumping CV Engineer: Detection, Tracking & 2D Calib

AIToolboard; United States; On-site; USD 82,656 - 165,312; Full time

Job summary

AIToolboard seeks a Computer Vision Engineer to build a production-grade perception stack for show-jumping footage. You will develop detection, tracking, obstacle understanding, and 2D metric calibration to output structured, metric-aware results.

Your work includes handling video ingestion, robust horse-and-rider detection, persistent tracking, obstacle segmentation, and 2D camera-to-ground mapping to enable accurate metric estimates like distances and speeds.

Qualifications

• Experience applying computer vision to video/sports footage.
• Strong background in object detection and multi-object tracking.
• Solid understanding of image-space geometry and camera perspective.
• Experience with 2D metric calibration using planar homography and RANSAC.

Responsibilities

• Ingest and preprocess videos (codecs, resolutions, FPS, orientation, stabilization).
• Detect horse and rider with state-of-the-art detectors (YOLO/RT-DETR/Detectron2).
• Maintain persistent tracking across frames (ByteTrack/BoT-SORT/DeepSORT).
• Detect obstacles and perform scene understanding for show-jumping arenas.
• Associate obstacles with jumps and determine the jumped obstacle.
• Segment full rounds into individual jump clips (per-obstacle).
• Reconstruct 2D horse trajectories with stable, low-jitter paths.

Skills

Python

OpenCV

PyTorch

YOLO

Detectron2

RT-DETR

Multi-object tracking

Segmentation

Homography

RANSAC

2D metric calibration

Image-space geometry

Pixel-to-meter conversion

Tools

Mask R-CNN

YOLO-Seg

SAM

PyTorch

Job description

AIToolboard seeks a Computer Vision Engineer to build a production-grade perception stack for show-jumping footage. You will develop detection, tracking, obstacle understanding, and 2D metric calibration to output structured, metric-aware results.

Your work includes handling video ingestion, robust horse-and-rider detection, persistent tracking, obstacle segmentation, and 2D camera-to-ground mapping to enable accurate metric estimates like distances and speeds.