Deployment
Edge devices
Raspberry Pi
You can run wasmVision natively on Raspberry Pi 3+ devices that use a 64-bit Linux-based operating system such as Raspberry Pi OS.
The wasmVision ARM64 builds already support hardware acceleration using NEON.
To install wasmVision, download the latest “wasmvision-linux-arm64.tar.gz” file with the release build, expand the compressed file, and put on the desired device. There are no external dependencies, so it should “just work”.
https://github.com/wasmvision/wasmvision/releases
NVIDIA Jetson
Native support for CUDA on the NVIDIA Jetson is coming soon.
AMD64
You can run wasmVision natively on AMD64 devices that use a 64-bit Linux-based operating system.
The wasmVision AMD64 builds already support hardware acceleration using SIMD.
To install wasmVision, download the latest “wasmvision-linux-amd64.tar.gz” file with the release build, expand the compressed file, and put on the desired device. There are no external dependencies, so it should “just work”.
https://github.com/wasmvision/wasmvision/releases
AMD64 with NVIDIA GPU
You can run the wasmVision container with CUDA support on AMD64 devices that use a 64-bit Linux-based operating system.
For more information, please see CUDA
Cloud
Google Cloud Run jobs
You can use Google Cloud Run jobs to run wasmVision on demand to process images or videos.
A typical setup would first create a container image based on the wasmVision Docker container image in the Google Cloud Artifact Registry.
Once the image was available in the Artifact Registry, create a new Google Cloud Job.
Execute the job with the specific parameters to process each image or video file.
You can also use Google Cloud storage volume mounts to make available a location to read or write the image/video data.
Microsoft Azure Container Apps jobs
You can use Microsoft Azure Container Apps jobs to run wasmVision on demand to process images or videos.
One way is to trigger a manual job that processes each image or video file.
Other cloud environments
More information on other cloud deployment scenarios is coming soon. Stay tuned!