https://kotlinlang.org logo
#ai
Title
# ai
v

Vivekpanchal64

04/20/2022, 9:49 AM
guys i have developed an app to capture video from road which i wanted to continuosly store and upload to server . but the device gets heatup too much because we are using openCv to detect optical flow and Ai model to detect vehicles. any idea how to reduce the over heat issue and make this process smooth. currently using camera 2 API.
j

James Black

04/21/2022, 3:57 AM
I am assuming you are doing this on Android, so you are using tensorflow lite? How many fps are you doing, or trying to do? Are you using YOLO or some similar approach to draw bounding boxes? Off the top of my head I would think see how slow you can reduce the framerate until it becomes an issue, and you may want to save the raw data to a video so you can send it to a server to use a better model. You can also use a simpler model on the phone.
v

Vivekpanchal64

04/21/2022, 4:01 AM
yes using tensorflow lite model. but the problem is when we continuously capture/record video the device gets too much heatup.
j

James Black

04/22/2022, 3:40 AM
I expect the frame rate is too high. You can limit the frame rate so perhaps 3 frames per second and see how that goes. Just keep playing with it and find a good number. Is tensor flow running on the GPU or CPU? https://www.tensorflow.org/lite/performance/gpu_advanced
v

Vivekpanchal64

04/22/2022, 3:41 AM
hmm thanks i'll try that. but might possible then i would losse alot of frames.
53 Views