What is the training dataset for the DeepStream sample model?

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide complete information as applicable to your setup.

Jetson Nano /DeepStream 4.0.1 /JetPack -4.2.2
CUDA(10.0) /TensorRT(5.1+) / OpenCV(3.3.1) /VisionWorks(1.6)

I want to compare the sample model with my custom model.
So, I want to know the information of Sample model (Primary_Detector_Nano/resnet10.caffemodel).
What dataset of sample model (resnet10) was used?
In addition, I would like to know the information of Primary_Detector_Nano / resnet10.caffemodel.

3

Hi,

This is our internal model which has similar architecture as DetectNet here:
https://devblogs.nvidia.com/detectnet-deep-neural-network-object-detection-digits/

The model is trained on our internal dataset so it is not available for the public user.
Thanks.

4

Hi AastaLLL

Thank you for answer.

6

Hello @AastaLLL, I have a related question about model normalization: in the config files provided with the model I found only the normalization parameter net-scale-factor but did not find any offset parameter for the mean values, offsets or mean-file. Does the dataset you used to train the resnet10 model have a zero pixel mean for all channels? This seems to me highly unlikely. Or else how does nvinfer behave if these values are not set?

7

Hi janos.tolgyesi,

Please help to open a new topic. Thanks