Video file test

[hirwa145]

Is there a way i can use video file as input instead of image/photo.

[jishang1602]

yes

[hirwa145]

how can i deploy it using deepstream?

[jishang1602]

how can i deploy it using deepstream? --> Not yet build.

Is there a way i can use video file as input instead of image/photo.? --> in main.cpp changing the 45th line from 0 to video.mp4

44. // init opencv stuff
45. VideoStreamer videoStreamer = VideoStreamer("video.mp4", videoFrameWidth, videoFrameHeight);
46. cv::Mat frame;

I hope i answered your query. Thanks

[hirwa145]

Thank you. I really appreciate your help and time.

[hirwa145]

I tried the solution you provided for using mp4 vido file as input. But i am getting errors

[jishang1602]

Can you please share the error

[hirwa145]

End generating TensorRT runtime models Parsing Directory: ../imgs Empty frame! Exiting... Try restarting nvargus-daemon by doing: sudo systemctl restart nvargus-daemon Counted 0 frames in 0 seconds! This equals nanfps.

[hirwa145]

After performing this command /mtcnn_facenet_cpp_tensorRT I didsudo systemctl restart nvargus-daemon` but i am still getting that error, what could be the reason?

I am not trying to compare from camera, just using mp4 video as input test

[jishang1602]

Did you put your database in the imgs folder? Also need to resize the video file to 640x480 size. Please do this changes and tell me the results

[hirwa145]

"Did you put your database in the imgs folder", you mean the image of known person to recognise in the video?

[jishang1602]

Yes the image database that you want to recognise in the video

[jishang1602]

This folder contains images of people you would like to recognize. The format of picture should class_name.jpg.

[hirwa145]

is it possible to use 1080x1920? by setting this to

34 int videoFrameWidth = 1080; 35 int videoFrameHeight = 1920

[hirwa145]

Thank you, i managed to load the video successfully, even tho it is showing new person on the face of a person whose face is included in the imgs file

[hirwa145]

Whilest doing the test, i realised that smaller faces are not detected, is there a way to solve this?

[jishang1602]

40 is the minimum pixel size for face detection. You can reduce it. But normally people are taking the smaller size upto 40 pixels

[hirwa145]

how can i change the pixel size? i mtcnn.cpp

//set minimal face size (weidth in pixels) nt minsize = 60; this side? and set it to lower size?