Hi
I have followed the MobileSSD SNPE tutorial, and have successfully converted the MobilneSSD 1.12 pretrained model into DLC and running SNPE DLC inference on Linux.
However, the problem occurred when I trained my oen MobileSSD model.
I am still able to convertt into DLC, execute the inference, but the results are different than the TF inference result.
I've followed the insturctions in the documentation tensorflow object_detection(https://github.com/tensorflow/models/tree/master/research/object_detection), train my ssd_mobilenet_v2, and then transfer to inference graph.
Confirm that the inference graph is ol, the score is between 0 and 1.
[0.7235, 0.6781, 0.4697, 0, 0, ...]
Folloewd the instructions in the documentation to convert the inference graph to dlc by the following command
snpe-tensorflow-to-dlc
--input_network fronze_inference_graph.pb
--input_dim Preprocessor/sub 1,300,300,3
--out_node detection_classes
--out_node detection_boxes
--out_node detection_scores
--output_path mobilenet_ssd.dlc
--allow_unconsumed_nodes
Run the dlc model.
snpe-net-run
--container mobilenet_ssd.dlc
--input_list raw_list.txt
raw_list.txt
# Postprocessor/BatchMultiClassNonMaxSuppression add
image.raw
View the detection_scores of model results:
[1.124, 1.17, 1.16, 1.1581, 1.15771,1.15596, 1.15431, 1.15336,1.15005,...]
The results of dlc is different of the tf model and number should be between 0 and 1.
Any suggestions on what I'm doing wrong?
My package version
Tensorflow version: 1.12.0
Snpe version:1.41.0
Thank for help.