Data Collector Client
Data collector (bn_data_collector
) is useful if you want to store the snapshot and inference results for data analysis.
Steps to enable data collector:
-
Prepare config file including the topic where AI engine will send inference results to:
For the default detection engine, you can create a
data_collector.json
{ "berrynet/engine/tflitedetector/result": "self.update" }
-
Run data collector:
$ bn_data_collector --data-dirpath /tmp/bn-outdir --topic-config data_collector.json
-
Check collected results in the indicated directory:
Every inference result will be saved to two files:
$ ls /tmp/bn-outdir 2019-05-26T13:33:52.277898.jpg 2019-05-26T13:33:52.277898.json ...
- image, w/ or w/o inference result overlay, depending on the output parameter set to AI engine.
- text in JSON format, w/o image string in base64 format.