gRPC Object Detection Server

gRPC server for Object Detection service. This is a personal project of Python implementation of a gRPC service with support for object detection from image and video / camera stream. For object detection, Ultralytics YOLO 🚀 is utilized, by default pretrained YOLOv8 (coco dataset) model is used. Custom trained model can also be used.

Installation

---

Use Python version 3.10

Conda Environment

Make sure to have anaconda or miniconda installed

1. conda create -n grpcenv python=3.10
2. conda activate grpcenv
3. cd grpc-ml-server
4. pip install -r requirements.txt

Virtualenv

1. cd grpc-ml-server
2. pip3 install virtualenv (if virtualenv is not installed)
3. python3 -m virtualenv venv
4. source venv/bin/activate
5. pip install -r requirements.txt

---

Additional

Install PyTorch with CUDA support

1. Follow the CUDA Installation Guide (Install cuda-11.7)

2.  pip install torch==1.13.0+cu{CUDA VERSION} torchvision==0.14.0+cu{CUDA VERSION} --extra-index-url https://download.pytorch.org/whl/cu{CUDA VERSION}
   

For system without CUDA

1.  pip install torch==1.13.0+cpu torchvision==0.14.0+cpu --extra-index-url https://download.pytorch.org/whl/cpu
   

Running the server

---

Compile proto file

When adding a new service, proto file should be compiled with the following command: bash compile_proto.sh <SERVICE_NAME>

Run server

python server.py --host <HOST> --port <PORT>
[OPTIONAL] --max-workers <MAX_NUM_WORKERS> --model-path <PATH_TO_CUSTOM_MODEL> --model-url <DOWNLOAD_URL>

TODO

---

• Implement CORS support for handling request from frontend
• Add method for multi-image request (unary-unary / stream-unary)
• Add environment variables
• Add docker support for deployment

Further improvements

---

• Add support for custom model integration
• Add service for instance / semantic segmentation, object tracking etc.

Contributions are welcome 😃