Automated species detection from camera traps

wkcmark

2020/01/30 08:43

Hello. I'm looking into the possibility to automatically identify species from camera traps.

This is for work based in the Middle East, mainly the United Arab Emirates and Saudi Arabia.

I'm new here and was wondering if anyone can point me in the right direction to get started?

I have familiarity with Python.

Musab

2020/03/24 09:50

Hello,

maybe I can help. Can you give me some more details about the project? It would be great if you have sample footage to train and validate

wkcmark

2020/03/24 10:01

In reply to Musab

Hello Musab. Thanks for responding. This is more at the enquiry stage at the moment. I work with a team of field ecologists and am looking to automate species detection when they are reviewing their camera trap footage.

Musab

2020/03/25 08:56

Something similar to this ?

https://youtu.be/K9a6mGNmhbc

wkcmark

2020/03/25 11:56

Wow thats really amazing work but not quite what I'm after.

What I want to be able to do is take hundreds of individual images from camera traps. Loop through all the images and identify the species in each image.

Musab

2020/03/25 14:59

I see. Im interested and would like to help. I will need the images to train the network. As many as possible.
if you dont have them yet, try to find similar images preferably of the same species. I will use them to test the performance of the detection.

AlexEcoVIS

2020/03/25 18:49

I'm not familiar with camera traps, but there are a couple of options:

1) If the animals tend to cover most part of the image, then you can train a CNN classifier to distinguish between species (available with the keras-Tensorflow modules in Python)

2) If, however, the animals only cover a small part of the image (e.g. in the distance), it might be better to use an object detector (I've used YOLOv2 in the past for fish detection), which however is not that straightforward, especially with Python (I used MATLAB)

In any case, keras-Tensorflow classification with Python might be the most straightforward option for your goal. You should also certainly have a look at Google's Wildlife Insights platform which is specialized for species classification from camera trap images.

radek

2020/03/27 10:33

In reply to wkcmark

This can be done, happy to help :) But I think I need to understand the situation a little bit more.

Do you already have the data for training / inference? Do you have any example images with the species in them annotated? Say a still from the camera with a tiger and a csv file referencing that file and annotating that there is a tiger in the image?

Would you like someone to do the developing and training of the deep learning model for you? I work as an AI research engineer at the Earth Species project and I am also a part of a community of deep learning practitioners where we apply cutting edge research to various problems. Here you can check a little initiative I started a couple of days ago to teach people how to work with audio (there is a related forum thread but unfortunately it is in closed forums for the time being as it is associated with a course that is under way). My main point is this - if you have the data and would like someone to help you out on the modelling part, I can coordinate this.

Alternatively, if you cannot release the data, I can point you to materials that can get you started to carry out the work yourself.