Yolov5 Custom Object Detection (Logo Detection)

Custom Object Detection model trained and deployed using FastAPI

To run the Docker image

Step 1

Run these docker commands

$ docker pull ankitgodle/detect_logo_api
$ docker run -p 8000:8000 --name detect_logo ankitgodle/detect_logo_api

Step 2

After the app startup is complete

Open URL - http://0.0.0.0:8000/docs/ in your browser. If you get Error 502, try URL - http://127.0.0.1:8000/docs

Step 3

Use POST/ api/detect/logo to upload the image and get the inferences

To run the FastAPI locally

Clone repo, install requirements.txt in a Python>=3.9.0 environment and follow step 2 and step 3 from above

$ git clone https://github.com/AnkitGodle/Logo_Detect_API
$ cd Logo_Detect_API
$ pip install -r requirements.txt
$ uvicorn app.main:app --reload --host 0.0.0.0 --port 8000

General Information

app - Contains the trained model, FastAPI code

logo_data - Contains training/test data

Yolo_train.ipynb - Model training code