View source for Camera detect

==Outline==
This website outlines the work we have been doing to analyse images for objects.
==Want to use this package and trial it?==
We have installed this onto our server. Please navigate here and have some fun: <b><li>http://114.142.160.204/~bosl/image_detect/camera_analysis.php</li></b>
You can use this site to analyse up to five of your own photos at a time. You can also change the accuracy threshold.
<br><br>
Steps to recreate this process is as follows on a linux operating system (largely following https://pjreddie.com/darknet/yolo/):
<li>Clone the gihub repo - git clone https://github.com/pjreddie/darknet.git then cd darknet then make</li>
<li>Grab the pre-trained weights - wget https://pjreddie.com/media/files/yolov3.weights</li>
<li>Then begin to play - e.g. ./darknet detect cfg/yolov3.cfg yolov3.weights data/dog.jpg</li>
<br>

==Case study 1 - Troups Creek Duck Counting==

To ensure that CT stays with us forever, we needed to have a way to automatically count ducks (she previously manually counted 1 million cows for us from individual pictures...and we were worried that she would run away when we told her about counting ducks at Troups Creek). So we found this awesome program called YoloV3 and decided to see if this model (trained on a generically available database). 
<br><br>
Here are some of the results of the automated counting system. You will see quite good agreement between the actual counts and the automated system. Just two of many images.
<br>

[[File:Example bird counts.png]]

<br><br>
Here is also the correlation between the manual counted method vs the automated system. There are over 300 photos used in this plot.<br>
[[File:Correlation Ducks.png]]