I have built a fungus identification bot that aims to identify, with reasonable accuracy, a fungus from a photograph. The intention is that you can go for a walk, capture a fungus on your phone and quickly chat with FungiBot to see if it can ID it from the photo.
FungiBot uses machine learning, so the more it is trained, the better it gets. Fungibot has been trained on 28 species of fungus at the moment, the most common in Britain.
My ask, please visit FungiBot and have a go, see if you can identify a fungus and please upload one that you have confidently identified. It should be intuitive, simply tag and upload. Easily accessible on the phone.
Can you explain more about how the system works as many people are now interested in automated systems to identify wildlife. For example do the images get sent to microsoft and are they getting the information too? Also how this system differs from many others that might be linked to Google or others.
Good questions and privacy is certainly not to be taken lightly. From my perspective, this is a hobby and learning exercise, so I do not collect any personal information and simply use the images to train the classifier. I do not share them or publish them anywhere. You can find FungiBot privacy info here: http://www.fungibot.com/privacy. Bear in mind, iSpotNature is also storing all images and these are publicly available for anyone to view.
If you are concerned about submitting images that you have identified then just use the bot for identification. I can provide full transparency on where your image is and how I can remove it if needed. Happy to answer or clarify more if you need.
Privicy is one thing but I was interested in whether microsoft or google or whoever are also learning how to identify things if you are using their service. It might be a hobby for you but for them it is big business if they can deal with computer vision and object recognition. Basically how does fungibot work and does the information end up with one or more of these big companies if you are using their service. I am not saying whether this is good or bad, just asking the question.
They absolutely are improving their own classifiers by getting customers to use their classifiers, so yes, they benefit from us using their services. I guess it depends on what it is you are uploading, in my case, I do not mind if Fungibot contributes to a broader level of fungus identification and understanding - this type of knowledge is a good thing in my opinion. I can see however that there would be cases where this would not be a good thing and an invasion of privacy.
I have copied this snippet from the Microsoft privacy site: "Microsoft Cognitive Services help developers build apps that can better comprehend data, such as by detecting faces and objects in images, understanding spoken words or interpreting commands. We collect and use your data to provide the services, which includes improving and personalising your experiences.
Cognitive Services collect and use many types of data, such as images, audio files, video files or text, all of which are retained by Microsoft. Microsoft has implemented business and technical measures designed to help de-identify some data that Cognitive Services retain. De-identification will not anonymise data completely; for example, a person in a photograph could still be recognised by someone who knows that person."
