edge-video-services / ekya Goto Github PK

Hi, thank you very much for this great work.
The bellevue_pretrained_models.tar.gz in the section Running Ekya with Urban Traffic Dataset is missing, can you please provide this pre-trained model? Thank you!

Hi, thanks for the great work and open sourcing your code!
I have gone through the labels GDrive for Bellevue Traffic Video Dataset and noticed that the labels are made only for the first 10 minutes every hour. Is this because you need just the training dataset? Do you have the labels for the entire 24 hours for any of the camera angles?

Thanks in advance

Hello,
I'm running and studying this code as your work inspires me.

I want to evaluate Ekya using object detection tasks via TinyYOLOv3/SSDLite with MobileNet.
However, according to the README, only faster_rcnn_resnet101_coco is downloaded into ekya/object_detection_model/ and used to prepare the urban traffic/building dataset in advance (generating sample_list for object classification tasks, in my understanding).

Could you inform me of how to run using object detection tasks?

I can't install nvidia-cuda-mps
E: Unable to locate package. nvidia-cuda-mps

I don't know how to run your code. Such as Running Ekya with Urban Building Dataset. This scene is for vegas, but you need to enter bash _driver_profiling_mp4_golden_bellevue_.sh. Can I only use the scheduling method of profiling? How can I see the inference process? Your paper testing on classification and detection tasks, respectively. Does each dataset perform testing on both tasks?