I am a beginner in everything programming, AI, deep learning etc. I recently started an internship in conservation at an NGO and they want to be able to monitor black bears using camera trap pictures. With these pictures they would like to be able to identify specific individual (e.g., bear number 1,2,3,4…). Several other projects are attempting this with bears or other species, but the teams are composed of people who’ve been developing software for years, and that made me unsure of my abilities to do such a project. My brother suggested using fast.ai as it seems to be the easiet solution for a fae recognition problem, but here’s my question:
For someone like me with very minimal programming experience (I’ve only used R for statistics and only now discovering python, pytorch and fast.ai), do you think it is reasonable for me to take on this challenge of developing a model that will be able to recognize individual bears using camera trap images from our own data base of pictures (which is quite limited, we may have 10-20 individuals)?
From what I’ve seen in the first four sessions, the code to distinguish between bear SPECIES seem very simple, I’m just not sure how complex it will turn out to be to actually be able to tell the model to:
“1. First, look at a picture and find all the faces in it
2. Second, focus on each face and be able to understand that even if a face is turned in a weird direction or in bad lighting, it is still the same person.
3. Third, be able to pick out unique features of the face that you can use to tell it apart from other people— like how big the eyes are, how long the face is, etc.
4. Finally, compare the unique features of that face to all the people you already know to determine the person’s name.”
Taken from – https://medium.com/@ageitgey/machine-learning-is-fun-part-4-modern-face-recognition-with-deep-learning-c3cffc121d78 –
From what I understand, fast.ai uses existing libraries for HUMAN facial recognition, which makes it much easier for someone like me to take on this challenge, but again my worry is whether it is relatively easy to do with a species for which no library currently exists.
Thank you so much for all your help.
Simon