Update object detection example to use new scannertools api

Author: fpoms

examples/apps/object_detection_tensorflow/kernels.py

@@ -3,14 +3,13 @@
 import cv2
 import os
 import scannerpy
-import scannerpy.stdlib.util
 import pickle
 import visualization_utils as vis_util
 import tarfile
 
 from scannerpy import FrameType, DeviceType
-from scannerpy.stdlib import tensorflow
 from scannerpy.types import NumpyArrayFloat32
+from scannertools import tensorflow
 from typing import Tuple, Sequence
 from tqdm import tqdm
 
@@ -24,12 +23,12 @@
 category_index = vis_util.create_category_index(categories)
 
 def download_and_extract_model(url, local_path=None):
-    path = scannerpy.stdlib.util.download_temp_file(url, local_path)
+    path = scannerpy.util.download_temp_file(url, local_path)
     tar_file = tarfile.open(path)
     for f in tar_file.getmembers():
         file_name = os.path.basename(f.name)
         if 'frozen_inference_graph.pb' in file_name:
-            local_path = scannerpy.stdlib.util.temp_directory()
+            local_path = scannerpy.util.temp_directory()
             tar_file.extract(f, local_path)
             model_path = os.path.join(local_path, f.name)
             break
@@ -40,11 +39,13 @@ def download_and_extract_model(url, local_path=None):
                               batch=2)
 class ObjDetect(tensorflow.TensorFlowKernel):
     def __init__(self, config, dnn_url):
+        print('objdet', config)
+        print([d.id for d in config.devices])
         tensorflow.TensorFlowKernel.__init__(self, config)
         self.dnn_url = dnn_url
         self.model_name = dnn_url.rsplit('/', 1)[-1]
         self.local_model_path = os.path.join(
-            scannerpy.stdlib.util.temp_directory(),
+            scannerpy.util.temp_directory(),
             self.model_name.rsplit('.')[0],
             'frozen_inference_graph.pb')
 

examples/apps/object_detection_tensorflow/main.py

@@ -1,5 +1,5 @@
-from scannerpy import Client, DeviceType
-from scannerpy.storage import NamedVideoStream, PythonStream
+from scannertools.storage.python import PythonStream
+import scannerpy as sp
 import os
 import sys
 import math
@@ -21,10 +21,10 @@ def main():
     print('Detecting objects in movie {}'.format(movie_path))
     movie_name = os.path.splitext(os.path.basename(movie_path))[0]
 
-    sc = Client()
+    sc = sp.Client()
 
     stride = 1
-    input_stream = NamedVideoStream(sc, movie_name, path=movie_path)
+    input_stream = sp.NamedVideoStream(sc, movie_name, path=movie_path)
     frame = sc.io.Input([input_stream])
     strided_frame = sc.streams.Stride(frame, [stride])
 
@@ -33,12 +33,14 @@ def main():
     objdet_frame = sc.ops.ObjDetect(
         frame=strided_frame,
         dnn_url=model_url,
-        device=DeviceType.GPU if sc.has_gpu() else DeviceType.CPU,
+        device=sp.DeviceType.GPU if sc.has_gpu() else sp.DeviceType.CPU,
         batch=2)
 
-    detect_stream = NamedVideoStream(sc, movie_name + '_detect')
+    detect_stream = sp.NamedVideoStream(sc, movie_name + '_detect')
     output_op = sc.io.Output(objdet_frame, [detect_stream])
-    sc.run(output_op)
+    sc.run(output_op,
+           sp.PerfParams.estimate(),
+           cache_mode=sp.CacheMode.Overwrite)
 
     print('Extracting data from Scanner output...')
     # bundled_data_list is a list of bundled_data
@@ -58,9 +60,11 @@ def main():
     drawn_frame = sc.ops.TFDrawBoxes(frame=strided_frame,
                                      bundled_data=bundled_data,
                                      min_score_thresh=0.5)
-    drawn_stream = NamedVideoStream(sc, movie_name + '_drawn_frames')
+    drawn_stream = sp.NamedVideoStream(sc, movie_name + '_drawn_frames')
     output_op = sc.io.Output(drawn_frame, [drawn_stream])
-    sc.run(output_op)
+    sc.run(output_op,
+           sp.PerfParams.estimate(),
+           cache_mode=sp.CacheMode.Overwrite)
 
     drawn_stream.save_mp4(movie_name + '_obj_detect')
 
@@ -73,5 +77,3 @@ def main():
 
 if __name__ == '__main__':
     main()
-
-