object-detector

This project implements basic object detection framework using python.

References for building this project is pyimagesearch.

1. Quick Start with default configuration file

1) Download dataset

• Download Caltech-101
• Download houses : This dataset will be used as the negative images

2) Extract dataset files

• Extract dataset files.
• Locate dataset directory as the following structure.

|--- [Project Directory]
|--- [datasets]
        |--- [caltech101]
                |--- [101_ObjectCategories]
                |--- [Annotations]
        |--- [houses]

3) Confirm the path of the dataset

Run "0_check_dataset_path.py". If you can find the following message, you can go to the next step.

c:\object-detector>python 0_check_dataset_path.py
Positive dataset location is correct
Negative dataset location is correct

4) Run the main driver

Run "main.py". It will automatically build object-detector with the 7-step framework.

4.1) Displayed Test Image

After finishing step 6, you can see the following test image and recognized bounding box.

In the case of using faces.conf we can see the following test image and recognized bounding box.

4.2) Average-Precision Evaluation

AP(Average Precision) is a measure of evaluating object detector. After finishing step 7, object detector's average precision score will be printed like this

Average Precision : 0.937441470843

It is also displayed precision-recall curve.