SSD Inception-V2
  • 25 Apr 2022
  • 1 Minute to read
  • Contributors
  • Dark
    Light

SSD Inception-V2

  • Dark
    Light

NetworkTensorRTOpenVINOOnnxRuntimeTranslator PluginMinimun required EVAReference link

SSD Inception-V2 

Tested

Not test, possible to optimize

Not Possible to run

adtrans_ssd

2.0+

sampleUffSSD

Original model : sampleUffSSD

TensorRT convert :

Follow the 

https://github.com/NVIDIA/TensorRT/tree/master/samples/opensource/sampleUffSSD

To convert Tensorflow model to UFF format

  • FP32 batchsize =1  

        /usr/src/tensorrt/bin/trtexec --uff=sample_ssd_relu6.uff --output=NMS --uffInput=Input,3,300,300 --buildOnly --saveEngine=YOUR_MODEL_NAME.engine --maxBatch=1

  • FP16 batchsize =1  

       /usr/src/tensorrt/bin/trtexec --uff=sample_ssd_relu6.uff --output=NMS --uffInput=Input,3,300,300 --buildOnly --saveEngine=YOUR_MODEL_NAME.engine --maxBatch=1 --fp16 

Sample command line launch for TensorRT :

gst-launch-1.0 filesrc location=street.mp4 ! decodebin ! videoconvert ! adrt model=ssdv2-b1.engine scale=0.0079 rgbconv=True ! adtrans_ssd max-count=100 label="files/ssd_coco_labels.txt" ! admetadrawer ! videoconvert ! ximagesink

Explanation of some plugins parameters :

adrt model=ssdv2-b1.engine scale=0.0079 rgbconv=True
  • model=ssdv2-b1.engine : Path to the location of the model used by plugin to inference
  • scale=0.0079 : Scale to be multiply with pixel RGB values to normalize pixel values to desired range. 0.0079 mean convert the input from range of (0 ~ 255) to (-1 ~ 1)
  • rgbconv=True : Convert the input image to RGB format from default BGR if set to True. False keep it as BGR



Was this article helpful?

What's Next