search

EdjeElectronics / TensorFlow-Object-Detection-API-Tutorial-Train-Multiple-Objects-Windows-10

How to train a TensorFlow Object Detection Classifier for multiple object detection on Windows
Apache License 2.0
2.92k stars 1.3k forks source link

Lot of lag while detecting Objects in Video file, Output is around only 2 Frame per second. #55

Open jayeshBaviskar opened 6 years ago

jayeshBaviskar commented 6 years ago

I have followed the exact steps of the tutorial but the output which I am getting is only at around 2 Frame Per second. I am Executing the code on below-mentioned PC Configuration:

Processor: Intel Xeon Processor - 3.60 GHz RAM: 64 GB Ram GPU: GeForce GTX 1060 6GB

Input Video File Format: MP4 Input Video File Resolution: 1920 X 1080 Input Video File Length: 2Min 45 sec

noelcodes commented 6 years ago

Sound like tensorflow-gpu isn't installed properly. Run model again with MSI afterburner hardware monitor running in background. I used this to check whether the model is using my GPU, if it is, usage % should be at least 30% or more. If its just hover at 0%, then you should re-install tensorflow-gpu + cuda + cudnn again.

MSI afterburner: https://www.msi.com/page/afterburner

jayeshBaviskar commented 6 years ago

I followed your steps, I Downloaded the MSI afterburner and verified that the TensorFlow is running on GPU, not on CPU.

Screenshot Attached.

gpu

karanmartian commented 6 years ago

same issue, is there any solution. i m getting about 5 fps

jayeshBaviskar commented 6 years ago

No, it works well with webcam feed, but on Video file it gives only around 3-5 FPS output. I think it is because the Processor uses its power to decode the video.

karanmartian commented 6 years ago

ok could you share your specs? i have gforce 940m 2gb, i5 processor, tf gpu. and webcam is working really slow for this, not sure why.