IDEA & APPLICATION using peekingduck’s computer vision to develop a functional system to detect potential high risk encounters between wild monkeys and park goers in Singapore, presenting such occurrances via a complementary front end to provide operators live coverage, expediting the process to activate emergency aid should the need arise.
TO SOLVE
- getting the machine to recognise difference between human and monkey
- defining a “safe distance”
- showing alert on front end
CONSIDERATIONS & MITIGATION
- security
- no storage of video data
- operators do not have open access to video feeds other than high risk occurrances
- AI is the primary touch point for data
- location of cameras
- areas of higher risk of monkey attacks (e.g. benches, rest areas, places where ppl are more likely to stop and eat
- along trail with known monkey traffic
Attached is the Google Colab link to our code: https://colab.research.google.com/drive/1AK6UU5ntd9xX6KgN7gwrWz-2HiYMJwhr?usp=sharing. Run it accordingly and you should be able to run and evaluate our model.
This links brings you to a Google Drive that containsthe weights and the sample photos used: https://drive.google.com/drive/folders/13S71cRKVUuH3sR1ZPuKtMSXmWC3L_KaY