ramsafin

# syntax=docker/dockerfile:1.7

ARG NVIDIA_CUDA_VERSION=13.1.1
ARG UBUNTU_VERSION=24.04

# ----------------------------
# Builder
# ----------------------------
FROM nvidia/cuda:${NVIDIA_CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION} AS builder

ramsafin

# syntax=docker/dockerfile:1.7

ARG TORCH_IMAGE_TAG=2.10.0-cuda13.0-cudnn9

# --- Builder ---
FROM pytorch/pytorch:${TORCH_IMAGE_TAG}-devel AS builder

ENV DEBIAN_FRONTEND=noninteractive \
    PIP_BREAK_SYSTEM_PACKAGES=1 \
    PIP_ROOT_USER_ACTION=ignore \

ramsafin

COLMAP Usage Guide

A guide to running COLMAP for structure-from-motion (SfM) and multi-view stereo (MVS).

Tested on: COLMAP 3.13 (CUDA, ceres-solver cuDSS).

Reference: https://colmap.github.io/

Overview

ramsafin

#!/usr/bin/env python3
from __future__ import annotations

import argparse
import math
from pathlib import Path
from typing import List, Tuple

import numpy as np
from PIL import Image

ramsafin

CUDA with Docker on WSL2

This guide covers:

• Installing CUDA Toolkit inside Ubuntu (WSL2) for native builds (nvcc, headers, libs)
• Enabling GPU in Docker containers on WSL2 (two options)

Overview

[!IMPORTANT]

ramsafin

Ceres Solver with CUDA + cuDSS

A guide to building Ceres Solver with CUDA support and NVIDIA's cuDSS direct sparse solver on WSL2/Ubuntu.

Tested on WSL2 (6.6.87.2-microsoft-standard-WSL2) with Ubuntu 24.04 LTS.

Prerequisites

Before proceeding, verify your GPU setup:

ramsafin

"""
COLMAP converter
- exhaustive matcher
- GPU auto-detection

Tested against:
  COLMAP 3.13.0 (CUDA)
"""

from __future__ import annotations

ramsafin

1. Install Miniconda

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
bash Miniconda3-latest-Linux-x86_64.sh

~/miniconda3/bin/conda init