Python Pipeline Testing Expert
Budget: $30 – $250 USD
IWork Zone Digital Twin — Pipeline Testing, Validation & Bug Fixes
Project Overview
We have a completed 11-stage Python pipeline that converts dashcam video and GPS/IMU telemetry into a top-down map of highway work zone assets (cones, drums, signs). The pipeline is built and running end-to-end. We need an engineer to test it thoroughly, validate outputs, and fix remaining issues.
Pipeline Summary
11-stage offline pipeline:
Stage 1-3: Telemetry preparation, keyframe extraction, video/GPS sync
Stage 4: COLMAP sparse SfM reconstruction
Stage 5: Metric scale alignment (GPS + physical reference measurement)
Stage 6-7: Manual and auto annotation tools (browser-based)
Stage 8: Multiview triangulation of asset 3D positions
Stage 9: Taper ordering and work zone measurements
Stage 10: Ground truth validation
Stage 11: Report generation (topdown.png, summary.md, evidence.csv)
Tech Stack
Python 3.10+
COLMAP (custom GPU build)
RT-DETR object detection model (HuggingFace)
NumPy, pandas, OpenCV, matplotlib
ENU coordinate system, GPS/IMU sensor fusion
What We Need
Testing — run the full pipeline on multiple video segments and verify outputs are correct at each stage
Validation — compare pipeline measurements (cone spacing, taper length) against known ground truth values, document accuracy
Bug Fixes — identify and fix remaining issues including:
Stage 9 taper/lane closure measurement accuracy
Asset deduplication and clustering logic
Scale factor computation for new recordings
Category mapping between detector output and pipeline schema
Documentation — document any changes made and update the existing codebase comments
What You'll Receive
Full codebase on private GitHub repo
Detailed handoff documentation (HANDOFF.md)
Sample video recordings with telemetry
Fine-tuned RT-DETR model for work zone detection
Existing test results for reference
Ideal Candidate
Strong Python debugging skills
Familiar with computer vision pipelines
Experience with SfM / COLMAP a plus
GPS/IMU sensor data experience a plus
Can read and understand an existing codebase quickly
Detail-oriented — comfortable with numerical validation
Project Overview
We have a completed 11-stage Python pipeline that converts dashcam video and GPS/IMU telemetry into a top-down map of highway work zone assets (cones, drums, signs). The pipeline is built and running end-to-end. We need an engineer to test it thoroughly, validate outputs, and fix remaining issues.
Pipeline Summary
11-stage offline pipeline:
Stage 1-3: Telemetry preparation, keyframe extraction, video/GPS sync
Stage 4: COLMAP sparse SfM reconstruction
Stage 5: Metric scale alignment (GPS + physical reference measurement)
Stage 6-7: Manual and auto annotation tools (browser-based)
Stage 8: Multiview triangulation of asset 3D positions
Stage 9: Taper ordering and work zone measurements
Stage 10: Ground truth validation
Stage 11: Report generation (topdown.png, summary.md, evidence.csv)
Tech Stack
Python 3.10+
COLMAP (custom GPU build)
RT-DETR object detection model (HuggingFace)
NumPy, pandas, OpenCV, matplotlib
ENU coordinate system, GPS/IMU sensor fusion
What We Need
Testing — run the full pipeline on multiple video segments and verify outputs are correct at each stage
Validation — compare pipeline measurements (cone spacing, taper length) against known ground truth values, document accuracy
Bug Fixes — identify and fix remaining issues including:
Stage 9 taper/lane closure measurement accuracy
Asset deduplication and clustering logic
Scale factor computation for new recordings
Category mapping between detector output and pipeline schema
Documentation — document any changes made and update the existing codebase comments
What You'll Receive
Full codebase on private GitHub repo
Detailed handoff documentation (HANDOFF.md)
Sample video recordings with telemetry
Fine-tuned RT-DETR model for work zone detection
Existing test results for reference
Ideal Candidate
Strong Python debugging skills
Familiar with computer vision pipelines
Experience with SfM / COLMAP a plus
GPS/IMU sensor data experience a plus
Can read and understand an existing codebase quickly
Detail-oriented — comfortable with numerical validation
Related categories:
Python
Linux
Software Architecture
Test Automation
Debugging
Docker
Continuous Integration
Computer Vision
Automation
CI/CD