YOLO based application for object detection

We will have a implementaiton of YOLO model and deployment of this model.

Why is it important to control the population of animals? High number of animals in some area may lead to habitat abuse, starvation, and death. That is one of the reasons for ecologists to control population size. Ecologists control density of species population with use of camera traps. Camera trap takes a sequence of pictures after it is triggered by motion nearby. One way Ecologists may evaluate the number of animals is to analyze this sequence manually. Alternatively, they may apply an algorithm which will analyze images and predict the number of animals in given images. Thus, the goal of this project is to evaluate the number of animals in a sequence of images taken from Camera Traps.

Structure

1. In /app everything related to service that predict number of animals.
2. In /analysis you can find plots and code for running the model to get the plots
3. In /data you can store everything related to images and related metadata. For example processed_test, processed_train, files from kaggle and data which were produced by model.
4. In model.ipynb you can find all code related to running the code, preparing the data e.t.c.
5. In data.yaml file we store data to use it for training by yolo torch cli

Application

To start a server run these commands:

$ docker-compose build
$ docker-compose up

Test using Postman

Use the same settings as here.

Dataset: Kaggle competition