v0.2.8

Docker images

RunPod (original serverless handler)

# CUDA 12.5 · Ubuntu 24.04 (recommended)
docker pull ghcr.io/alx/colmap-serverless:v0.2.8
docker pull ghcr.io/alx/colmap-serverless:v0.2.8-cuda12.5

# CUDA 12.4 · Ubuntu 22.04 (older driver support)
docker pull ghcr.io/alx/colmap-serverless:v0.2.8-cuda12.4

Run locally (RunPod):

docker run --rm --gpus all ghcr.io/alx/colmap-serverless:v0.2.8 \
python3 handler.py --test_input '{"input":{"video_url":"<your-video-url>"}}'

Vast.ai (docker ENTRYPOINT mode)

# CUDA 12.5 · Ubuntu 24.04 (recommended)
docker pull ghcr.io/alx/colmap-serverless:v0.2.8-vastai
docker pull ghcr.io/alx/colmap-serverless:v0.2.8-vastai-cuda12.5

# CUDA 12.4 · Ubuntu 22.04 (older driver support)
docker pull ghcr.io/alx/colmap-serverless:v0.2.8-vastai-cuda12.4

Run locally (Vast.ai entrypoint):

docker run --rm --gpus all \
-e VIDEO_URL="<your-video-url>" \
-e NUM_FRAMES=150 \
-e MATCHING=sequential \
ghcr.io/alx/colmap-serverless:v0.2.8-vastai

Use the vastai tag in your Vast.ai template's Image Path:Tag field
and set VIDEO_URL (+ any overrides) as environment variables.

Full Changelog: v0.2.7...v0.2.8