[Tensorfilter] Need NN Model to test other/tensors
Opened this issue · 3 comments
Monday Jul 30, 2018 at 10:04 GMT
Originally opened as https://github.sec.samsung.net/STAR/nnstreamer/issues/366
Issue Description
May need Fast RCNN Model for object detection. The output of this RCNN is other/tensors.
- Input : input image
- Output : detection_boxes, detection_scores, detection_classes, num_detections
However, it is somewhat difficult to find pre-translated tflite model. According to the tflite manual, we could use toco to translate flatbuffer. It is worth to try. ( It is much better if we could find the proper tflite model from internet.)
Monday Jul 30, 2018 at 10:04 GMT
cibot: Thank you for posting issue #366. The person in charge will reply soon.
Monday Jul 30, 2018 at 11:02 GMT
I can build toco successfully on Ubuntu 16.04 with command below:
$ ./bazel-0.15.2-install-linux-x86_64.sh # install bazel > 0.15
$ bazel build tensorflow/contrib/lite/toco:toco
$ ./bazel-bin/tensorflow/contrib/lite/toco/toco --input_file=./data/frozen_inference_graph.pb --output_file=./data/frozen_inference_graph.tflite --input_format=TENSORFLOW_GRAPHDEF --output_format=TFLITE --inference_type=QUANTIZED_UINT8 --input_shape="1,640,480,3" --input_array='image_tensor' --output_array='detection_boxes','detection_scores','num_detections','detection_classes' --std_value=127.5 --mean_value=127.5But still there is some error. Need to do more.
Tuesday Jul 31, 2018 at 02:04 GMT
Actually, there is a model guide for object detection.
Though I got the result of the model, it doesn't work properly.
Below document can help you
Running on mobile with TensorFlow Lite