Website YOLO Detector

The repository stores the back-end for the Flask application, which serves the requests coming from the detector at my website.

How Does it Work

To put it simply, it receives a user image and runs an object detection algorithm on it (YOLO v3). Once the predictions are retrieved, the resulting image with bounding boxes is sent back to the front-end. In this README I provide the environment setup for the compute machine. However, setting up the bach-end machine is just the tip of an iceberg. The whole engineering pipeline includes many other steps full of caveats. Just to scratch the surface, I undertook the following steps to build this project:

1. writen the website front-end (v-iashin/v-iashin.github.io)
2. obtained a domain name (Freenom — I wouldn't recommend it though!)
3. rented an instance and reserved an IP for it (GoogleCloud)
4. added DNS entries mapping my domain to the instance IP (Freenom again)
5. signed instance-side digital certificates for HTTPs for my domain (Let's Encrypt)
6. setted up the back-end environment on my instance + detector implementation (THIS repo)

If you are interested in the details of each step, let me know in Issues.

Setting up the Environment

Download the YOLOv3 weights

bash ./weights/download_weights_yolov3.sh

Install the conda environment

conda env create -f ./conda_env.yml
conda activate detector