BoquilaHUB

Run AI models to monitor and protect nature. Locally, no cloud.

List of AIs:

AI Name	Description	Input Type	Available?
boquilanet-gen	Generic animal detection	Image	✅
boquilanet-cl	Chilean fauna classification	Image	✅
boquilanet-eu	European fauna classification	Image	✅
megadetector v5a	MegaDetector (animals, vehicles, people)	Image	✅
boquila-fire	Wildfire detection	Image	✅
unnamed	Bird detection	Audio	On the way
unnamed	Chilean birds classification	Audio	On the way
unnamed	Automated marine acoustics	Audio	On the way
unnamed	Tiny LLM	Text	On the way

Image = Image files, video files, video feed.

List of Platforms

Platform	Production ready
Windows	✅
Android	On the way
Linux	On the way
MacOS	Not soon
iOS	Not soon

List of Runtimes

Runtime	Description	Requirements	Available?
cpu	Your average CPU	Having a CPU	✅
NVIDIA CUDA	CUDA execution provider for NVIDIA GPUs (Maxwell 7xx and above)	Requires CUDA v12.4 and cuDNN 8.9.2.26+	✅
NVIDIA TensorRT	TensorRT execution provider for NVIDIA GPUs (GeForce 9xx series and above)	Requires CUDA v11.4+ and TensorRT v8.4+	On the way
AMD ROCm	ROCm execution provider for AMD GPUs	Requires ROCm-supported AMD GPUs	On the way
AMD MIGraphX	MIGraphX execution provider for AMD GPUs	Requires AMD ROCm and MIGraphX	On the way
AMD Vitis AI	Vitis AI execution provider for Xilinx FPGA devices	Requires Xilinx Vitis AI software stack	On the way
Intel OpenVINO	OpenVINO execution provider for Intel Core CPUs (6th generation and above)		On the way
Intel oneDNN	Intel oneDNN execution provider for x86/x64 targets		On the way
Microsoft DirectML	DirectML execution provider for Windows x86/x64 targets with dedicated GPUs	Requires DirectX 12 support and dedicated GPUs	On the way
Microsoft Azure	Azure AI execution provider for cloud-based inference on Microsoft Azure	Requires Azure cloud environment	On the way
Qualcomm QNN	Qualcomm AI Engine Direct SDK execution provider for Qualcomm chipsets		On the way
Apple CoreML	CoreML execution provider for Apple devices	Requires macOS or iOS	On the way
XNNPACK	XNNPACK execution provider for optimized inference on ARM and x86 devices	Requires devices with SIMD instruction sets	On the way
Huawei CANN	Huawei CANN execution provider for Huawei Ascend AI processors	Requires Huawei Ascend chipsets	On the way
Android NNAPI	Android NNAPI execution provider for mobile devices with NNAPI support	Requires Android device with NNAPI	On the way
Apache TVM	Apache TVM execution provider for multiple backends	Requires compilation with TVM	On the way
Arm ACL	Arm Compute Library (ACL) execution provider for Arm devices	Requires devices with ARM processors	On the way
ArmNN	ArmNN execution provider for ARM-based devices	Requires devices with ARM processors	On the way
Rockchip RKNPU	Rockchip RKNPU execution provider for Rockchip NPUs	Requires devices with Rockchip NPUs	On the way

Requirements are gonna be clearer in the future.

Tech stack

We use:

• Flutter 3.19.2 for the UI

• Rust 1.79.0 for the inference pipeline

Open source?

In the future, we'll be fully open source. Right now, you can see the UI code.