Accuracy and Recall of pretrained model

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

I am selecting pretrained model(s) from TLT/Deepstream SDK. I noticed that benchmark only report the accuracy and no recall score, for example:
Model Name / Network Architecture / Number of classes / Accuracy / Use Case
TrafficCamNet / DetectNet_v2-ResNet18 / 4 / 83.5% / Detect and track cars

I am interested on the following pretrained models:

  1. TrafficCamNet (above)
  2. resnet10 (in deepstream sdk: model-engine-file=/opt/nvidia/deepstream/deepstream-5.0/samples/models/Primary_Detector/resnet10.caffemodel_b1_gpu0_int8.engine)

I would like to pick the higher accuracy as well as higher recall model for my application. I wonder where can I find the accuracy and recall numbers for these two models. Please advise. Thanks a lot.

3

Can you share your setup with us?

4

AGX Xavier
DeepStream 5.0
JetPack 4.4

5

Changing to TLT forum.

For the recall result of trafficcamnet, please see https://ngc.nvidia.com/catalog/models/nvidia:tlt_trafficcamnet

|Content|Precision|Recall|Accuracy|
|TrafficcamNet|92.65|89.95|83.9|

6

@Morganh

Thank you for the info. That’s the number I am looking for. Thanks.