eco_hackaton_amur_tigers
what is done in the end of day 1:
- Idea
- Models to tag photos
- Frontend draft
Descritption
The idea is as simple as that Stuff who are getting photos from camera-traps are very busy with analysing all data from them by hands. It is very time-consuming to look through all this data. Our team thought to automate process of spotting particular animal and then recognize problematic animals or puppets.
Problems we encountered
In the end of day one we discovered that actually for gender tagging and for weight tagging there is no data at all.
The fix and new concept
So the solution is to provide one who wants to analyse camera-traps data with an oppotunity to add his own tags and data for them. For each tag there will be a binary claassifier which is trained on google servers and by trained I mean not only it's weights but also the architecture, so it is only data which is needed to start and process of training. The simple interface is providing anyone with oppotunity to analyse almost arbitrary data for almost arbitrary attribute.
Final look
In the end, our product is a library of simple classifiers which are able to recognize particular animals and then add tags for it's photo if it is problematic (wounds or starving) or if there are а few more little tigers just appeared in this world. But if one wants to know something else about animals it is simple to add a new classifier and to tag your animals with a new property.
Further development
- identify unique tigers, there are only 540 after all.
Built With
- figma
- javascript
- tensorflow.js
- vuetify