Hi, I tried a few Caffe face detectors.
e.g. deploy.prototxt and res10_300x300_ssd_iter_140000_fp16.caffemodel from https://github.com/opencv/opencv/tree/master/samples/dnn/face_detector
All gave the same error as the following:
[libprotobuf ERROR google/protobuf/text_format.cc:298] Error parsing text-format ditcaffe.NetParameter: 1192:19: Message type “ditcaffe.LayerParameter” has no field named “prior_box_param”
It is because the layers is not supported?
Do anyone know how to resolve this issue? Thanks.
Did you find any solutions? I have the same problem?
gst-nvinfer plugin calls nvidia tensorRT to implement. There are a lot of layers like priorBox, detectionOutput, flatten, … which tensorRT does not support. You need to implement IPLUGIN for these layers. You can refer to Deepstream 3.0 DeepStream_Release/sources/ objectDetector_FasterRCNN and objectDetector_SSD
@ChrisDing thank you for the reply and I’m working on implementing them right now. Hope these kind of layer support will be there with the upcoming releases to make it easier running the application out of the box faster. Appreciate the hard work done on TensorRT example set.