Focoos pre-trained models

Model Name	Task	Metrics	Domain
focoos_object365	Detection	-	Common Objects, 365 classes
focoos_rtdetr	Detection	-	Common Objects, 80 classes
focoos_cts_medium	Semantic Segmentation	-	Autonomous driving, 30 classes
focoos_cts_large	Semantic Segmentation	-	Autonomous driving, 30 classes
focoos_ade_nano	Semantic Segmentation	-	Common Scenes, 150 classes
focoos_ade_small	Semantic Segmentation	-	Common Scenes, 150 classes
focoos_ade_medium	Semantic Segmentation	-	Common Scenes, 150 classes
focoos_ade_large	Semantic Segmentation	-	Common Scenes, 150 classes
focoos_aeroscapes	Semantic Segmentation	-	Drone Aerial Scenes, 11 classes
focoos_isaid_nano	Semantic Segmentation	-	Satellite Imagery, 15 classes
focoos_isaid_medium	Semantic Segmentation	-	Satellite Imagery, 15 classes

Focoos

Focoos is a comprehensive SDK designed for computer vision tasks such as object detection, semantic segmentation, instance segmentation, and more. It provides pre-trained models that can be easily integrated and customized by users for various applications. Focoos supports both cloud and local inference, and enables training on the cloud, making it a versatile tool for developers working in different domains, including autonomous driving, common scenes, drone aerial scenes, and satellite imagery.

Key Features

• Pre-trained Models: A wide range of pre-trained models for different tasks and domains.
• Cloud Inference: API to Focoos cloud inference.
• Cloud Training: Train custom models with the Focoos cloud.
• Multiple Local Inference Runtimes: Support for various inference runtimes including CPU, GPU, Torchscript CUDA, OnnxRuntime CUDA, and OnnxRuntime TensorRT.
• Model Monitoring: Monitor model performance and metrics.

🐍 Setup

We recommend using UV as a package manager and environment manager for a streamlined dependency management experience. Here’s how to create a new virtual environment with UV:

pip install uv
uv venv --python 3.12
source .venv/bin/activate

Focoos models support multiple inference runtimes. To keep the library lightweight and to allow users to use their environment, optional dependencies (e.g., torch, onnxruntime, tensorrt) are not installed by default. Foocoos is shipped with the following extras dependencies:

• [torch]: torchscript CUDA
• [cuda]: onnxruntime CUDA
• [tensorrt]: onnxruntime TensorRT

CPU only or Remote Usage

uv pip install focoos[cpu] git+https://github.com/FocoosAI/focoos.git

GPU Runtimes

Torchscript CUDA

uv pip install focoos[torch] git+https://github.com/FocoosAI/focoos.git

OnnxRuntime CUDA

uv pip install focoos[cuda] git+https://github.com/FocoosAI/focoos.gi

OnnxRuntime TensorRT

To perform inference using TensorRT, ensure you have TensorRT version 10.5 installed.

sudo apt-get install tensorrt

uv pip install focoos[tensorrt] git+https://github.com/FocoosAI/focoos.git

🤖 Cloud Inference

from focoos import Focoos

focoos = Focoos(api_key=os.getenv("FOCOOS_API_KEY"))

model = focoos.get_remote_model("focoos_object365")
detections = model.infer("./image.jpg", threshold=0.4)

🤖 Cloud Inference with Gradio

setup FOCOOS_API_KEY_GRADIO environment variable with your Focoos API key

uv pip install focoos[dev] git+https://github.com/FocoosAI/focoos.git

gradio gradio/app.py

Local Inference

from focoos import Focoos

focoos = Focoos(api_key=os.getenv("FOCOOS_API_KEY"))

model = focoos.get_local_model("focoos_object365")

detections = model.infer("./image.jpg", threshold=0.4)

Docker and devcontainers

For container support, Focoos offers four different Docker images:

• focoos-cpu: only CPU
• focoos-onnx: Includes ONNX support
• focoos-torch: Includes ONNX and Torchscript support
• focoos-tensorrt: Includes ONNX, Torchscript, and TensorRT support

This repository also includes a devcontainer configuration for each of the above images. You can launch these devcontainers in Visual Studio Code for a seamless development experience.