No description, website, or topics provided.
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
sample
utils
LICENSE
README.md
inference.py
webcam_inference.py
yolo_model.py

README.md

YOLO-Object-Detection optimization on Xeon scalable processors

In this project, optimization of TensorFlow code is performed for an object detection application to obtain real-time performance.
Please refer the following paper for all the details regarding performance optimizations,
https://colfaxresearch.com/yolo-optimization/

Rquirements:

Numpy
Python 2.7
Tensroflow
OpenCV

Steps to use this code:

  1. Go to utils/ and run:
    $ python config.py
    this downloads the darknet weight files. Also, fuses batchnorm layers and creates TensorFlow Ckpt files.

  2. To run image inference:
    $ python inference.py , to run TinyYolo model
    $ python inference.py --image= [image path]
    $ python infernce.py --v2 , to run YoloV2 model
    $ NUM_INTER_THREADS=2 NUM_INTRA_THREADS=8 python inference.py --par , to run parallel TensorFlow session(Inter/Intra op threads), if it is supported in your system.

  3. To run Webcam inference:
    $ python webcam_inference.py

Please refer the paper mentioned above to know more about the system used for testing and the versions of software tools used.