ambianic / fall-detection

Python ML library for people fall detection
Apache License 2.0
83 stars 16 forks source link

Implement more test cases based on public and user contributed video and image data of elderly falls #4

Open ivelin opened 3 years ago

ivelin commented 3 years ago

Is your feature request related to a problem? Please describe. We continuously receive images from users with examples of false positives or missed positives. This is expected iterative process of improvement.

Describe the solution you'd like We should add to the testsuite examples from publicly available elderly fall videos in order to improve the performance of our fall detection algorithm.

Here is one such dataset.

Additional context Attached are additional images reported from ambianic users with false positives. Screen Shot 2021-01-12 at 2 10 10 PM Screen Shot 2021-01-12 at 2 10 23 PM Screen Shot 2021-01-19 at 4 22 48 PM

ivelin commented 3 years ago

@bhavikapanara here is again the link to the data samples I extracted for our test suite. These are sequences of image falls of elderly from public video sources.

ivelin commented 3 years ago

@bhavikapanara more user contributed test images. These show the challenges that the MobileNet PoseNet model has detecting people poses from a camera mounted on a ceiling corner looking down towards the door and floor.

ivelin commented 3 years ago

@bhavikapanara Five public videos of elderly falls to consider for the test suite:

https://www.youtube.com/watch?v=p5i4z3sNaKM https://www.youtube.com/watch?v=1IsM08Sh_wg https://www.youtube.com/watch?v=3mDmkOxprN0 https://www.youtube.com/watch?v=0VqvZGhK1o8 https://www.youtube.com/watch?v=Od_RgAP8ojk