GreenLab
Hi,
I'm currently working on a solution to save snakes from humans by helping humans to identify whether a particular snake is venomous or not. The idea came into my mind when one of my friend told me that there are hundrads of thousands of non-venomous snakes are getting killed daily by humans because these snakes are wrongly identified as venoumous snakes.
My idea is to develop a mobile app that detects the snake using a camera app and let the user know a particular snake is venomous or not (and many other useful info). With such a solution, helping humans to take informed decisions, I hope we can save many reptiles, especially endangered and native snakes in a particular geographic region.
Currently I'm experimenting with machine learning algorithms to build a suitable model. I hope to use TensotFlow and Firebase ML Kit to build the AI model and Flutter to build the mobile app.
I need help in following;
1. Are there any similar solutions?
2. To train and test such a ML algorithm, I need may be more than 10,000 labeled images of different snake species. Is there such an image database available?
3. What are the other possible features such an app could have?
4. What are the other human-animal conflict scenarios such a solution could be helpful?
5. How to maintain such an app in long run, how to cover costs?
6. Are there any funding/resourcing orgs with similar interest?
7. Any other helpful ideas?
9 October 2019 8:45pm
Hi Nilaksha,
Interesting idea. My ideas are usually not that helpful, but I thought I would chime in with the following:
(1) Start small and local, this will help reduce the size of the training data set you will need. So if you can work with a specific set of communities in a particular area that has a known set of snake species, both venemous and non-venemous, you only need images for those species and you won't need as many.
(2) iNaturalist has a pretty decent database (not thousands, but hundreds) of identified images. I have no idea what the restrictions would be around getting access to the database, but this is a Citizen Science organization, so I don't THINK it should be too complicated, especially if your initial scope is limited.
(3) Venemous snakes need "love" too. PLEASE make sure the app does not encourage the locals to kill the venemous snakes unless absolutely necessary. While it is admirable to prevent the accidental killing of a non-venemous snake, you shoudl not be encouraging the purposeful killing of the venemous snakes, but should be encouraging people to stay clear if the snake is in the wild and get professional assistance if the snake somewhere that could be dangerous to humans or other animals.
(4) Look into organizations like the Rainforest Aliance, OpenAI, etc., which are non-profits doing work in or with AI, they might be able to help give you a leg up.
(5) If you start local, you could build into the app the contact information for the professionals to come in and deal with the dangerous snakes.
Hopefully some of this is useful.
Good luck,
Drue
Drue Freeman