search

shariqfarooq123 / AdaBins

Official implementation of Adabins: Depth Estimation using adaptive bins
GNU General Public License v3.0
730 stars 156 forks source link

Realtime Distance Calculation #34

Closed saiteja1904 closed 2 years ago

saiteja1904 commented 3 years ago

Hi I am trying to implement this model along with yolo Y5 ,realtime by using cameras on a motorbike to calculate the distance of vehicles,while riding the bike, but the accuracy is very bad is there any way to improve the accuracy ?,and is it possible to implement this on Jetson Nano ?

I3aer commented 3 years ago

Hi, There is a paper that examines "How do NNs see depth in single images?" published by ICCV. According to that paper, depth estimation networks cannot completely detect and account for changes in camera pose. Since depth estimation depends on camera pose, monocular depth estimation networks over-/under-estimates depths from images under change in camera pose. Unfortunately, you cannot obtain reliable results unless you fine-tune monocular depth estimation networks to your dataset.