Skip to content

Commit

Permalink
Merge pull request OpenDroneMap#1549 from NtskwK/patch-2
Browse files Browse the repository at this point in the history
feat: instructions of GPU acceleration
  • Loading branch information
pierotofy authored Sep 2, 2024
2 parents 4ab6eae + 3aba66f commit 4c089da
Showing 1 changed file with 5 additions and 2 deletions.
7 changes: 5 additions & 2 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -54,7 +54,7 @@ To run a standalone installation of WebODM (the user interface), including the p
* 100 GB free disk space
* 16 GB RAM

Don't expect to process more than a few hundred images with these specifications. To process larger datasets, add more RAM linearly to the number of images you want to process. A CPU with more cores will speed up processing, but can increase memory usage. GPU acceleration is also supported. To make use of your CUDA-compatible graphics card, make sure to pass `--gpu` when starting WebODM. You need the nvidia-docker installed in this case, see https://github.com/NVIDIA/nvidia-docker and https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#docker for information on docker/NVIDIA setup.
Don't expect to process more than a few hundred images with these specifications. To process larger datasets, add more RAM linearly to the number of images you want to process. A CPU with more cores will speed up processing, but can increase memory usage. GPU acceleration is also supported on Linux and WSL. To make use of your CUDA-compatible graphics card, make sure to pass `--gpu` when starting WebODM. You need the nvidia-docker installed in this case, see https://github.com/NVIDIA/nvidia-docker and https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#docker for information on docker/NVIDIA setup.

WebODM runs best on Linux, but works well on Windows and Mac too. If you are technically inclined, you can get WebODM to run natively on all three platforms.

Expand All @@ -77,9 +77,11 @@ To install WebODM manually on your machine with docker:
2. give Docker enough CPUs (default 2) and RAM (>4Gb, 16Gb better but leave some for Windows) by going to Settings -- Advanced

3. select where on your hard drive you want virtual hard drives to reside (Settings -- Advanced -- Images & Volumes).

4. If you want to run the processing component (NodeODM) with GPU acceleration, install [WSL](https://learn.microsoft.com/windows/wsl/) and [set up GPU acceleration](https://learn.microsoft.com/windows/wsl/tutorials/gpu-compute). It's supported on Windows 11 or Windows 10, version 21H2 or higher.

### Installation with Docker
* From the Docker Quickstart Terminal or Git Bash (Windows), or from the command line (Mac / Linux), type:
* From the Docker Quickstart Terminal or Git Bash (Windows), or from the command line (Mac / Linux / WSL), type:
```bash
git clone https://github.com/OpenDroneMap/WebODM --config core.autocrlf=input --depth 1
cd WebODM
Expand Down Expand Up @@ -183,6 +185,7 @@ Getting a `No space left on device` error, but hard drive has enough space left
Cannot start WebODM via `./webodm.sh start`, error messages are different at each retry | You could be running out of memory. Make sure you have enough RAM available. 2GB should be the recommended minimum, unless you know what you are doing
While running WebODM with Docker Toolbox (VirtualBox) you cannot access WebODM from another computer in the same network. | As Administrator, run `cmd.exe` and then type `"C:\Program Files\Oracle\VirtualBox\VBoxManage.exe" controlvm "default" natpf1 "rule-name,tcp,,8000,,8000"`
On Windows, the storage space shown on the WebODM diagnostic page is not the same as what is actually set in Docker's settings. | From Hyper-V Manager, right-click “DockerDesktopVM”, go to Edit Disk, then choose to expand the disk and match the maximum size to the settings specified in the docker settings. Upon making the changes, restart docker.
On Linux or WSL, Warning: `GPU use was requested, but no GPU has been found` | Run `nvidia-smi` (natively) or `docker run --rm --gpus all nvidia/cuda:11.2.2-devel-ubuntu20.04 nvidia-smi` (docker) to check with [NVIDIA driver](https://www.nvidia.com/drivers/unix/) and [NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html).

#### Images Missing from Lightning Assets

Expand Down

0 comments on commit 4c089da

Please sign in to comment.