kinhong / OpenLabeler

OpenLabeler is an open source desktop application for annotating objects for AI appplications
Apache License 2.0
112 stars 11 forks source link

TF error #16

Closed krzynio closed 2 years ago

krzynio commented 4 years ago

Hi,

I have finetuned the model but inference does not work. It was trained with your docker image (gpu).

2020-05-16 09:31:03.185350: I tensorflow/cc/saved_model/reader.cc:54] Reading meta graph with tags { serve } 2020-05-16 09:31:03.253808: I tensorflow/cc/saved_model/loader.cc:202] Restoring SavedModel bundle. 2020-05-16 09:31:03.434418: I tensorflow/cc/saved_model/loader.cc:151] Running initialization op on SavedModel bundle at path: /home/krzynio/Pobrane/ssd_mobilenet_v1_coco_2018_01_28/train/export/Servo/1589595393 2020-05-16 09:31:03.491558: I tensorflow/cc/saved_model/loader.cc:311] SavedModel load for tags { serve }; Status: success. Took 320378 microseconds. 2020-05-16 09:31:03 INFO com.easymobo.openlabeler.tensorflow.ObjectDetector: Loaded saved model from /home/krzynio/Pobrane/ssd_mobilenet_v1_coco_2018_01_28/train/export/Servo/1589595393 2020-05-16 09:31:03 INFO com.easymobo.openlabeler.tensorflow.ObjectDetector: MODEL SIGNATURE 2020-05-16 09:31:03 INFO com.easymobo.openlabeler.tensorflow.ObjectDetector: Inputs: 2020-05-16 09:31:03 INFO com.easymobo.openlabeler.tensorflow.ObjectDetector: 1 of 1: serialized_example (Node name in graph: tf_example:0 , type: DT_STRING) 2020-05-16 09:31:03 INFO com.easymobo.openlabeler.tensorflow.ObjectDetector: Outputs: 2020-05-16 09:31:03 INFO com.easymobo.openlabeler.tensorflow.ObjectDetector: 1 of 4: detection_classes (Node name in graph: detection_classes:0 , type: DT_FLOAT) 2020-05-16 09:31:03 INFO com.easymobo.openlabeler.tensorflow.ObjectDetector: 2 of 4: num_detections (Node name in graph: num_detections:0 , type: DT_FLOAT) 2020-05-16 09:31:03 INFO com.easymobo.openlabeler.tensorflow.ObjectDetector: 3 of 4: detection_boxes (Node name in graph: detection_boxes:0 , type: DT_FLOAT) 2020-05-16 09:31:03 INFO com.easymobo.openlabeler.tensorflow.ObjectDetector: 4 of 4: detection_scores (Node name in graph: detection_scores:0 , type: DT_FLOAT) 2020-05-16 09:31:03 INFO com.easymobo.openlabeler.tensorflow.ObjectDetector: ----------------------------------------------- 2020-05-16 09:31:06 WARNING com.easymobo.openlabeler.ui.TagBoard: Fail to detect java.lang.IllegalArgumentException: No Operation named [] in the Graph at org.tensorflow.Session$Runner.operationByName(Session.java:384) at org.tensorflow.Session$Runner.parseOutput(Session.java:393) at org.tensorflow.Session$Runner.feed(Session.java:132) at com.easymobo.openlabeler.tensorflow.ObjectDetector.detect(ObjectDetector.java:170) at com.easymobo.openlabeler.ui.TagBoard.lambda$findHints$36(TagBoard.java:500) at java.base/java.lang.Thread.run(Thread.java:834)

2020-05-16 09:31:07 WARNING com.easymobo.openlabeler.ui.TagBoard: Fail to detect java.lang.IllegalArgumentException: No Operation named [] in the Graph at org.tensorflow.Session$Runner.operationByName(Session.java:384) at org.tensorflow.Session$Runner.parseOutput(Session.java:393) at org.tensorflow.Session$Runner.feed(Session.java:132) at com.easymobo.openlabeler.tensorflow.ObjectDetector.detect(ObjectDetector.java:170) at com.easymobo.openlabeler.ui.TagBoard.lambda$findHints$36(TagBoard.java:500) at java.base/java.lang.Thread.run(Thread.java:834)

kinhong commented 4 years ago

@krzynio - from the log entries, it seems that the model is missing the image_tensor or the encoded_image_string_tensor input operations that OpenLabeler requires to perform object detection. Also, the output operation: detection_scores is missing. Here is a typical model signature that works: MODEL SIGNATURE Inputs: 1 of 1: inputs (Node name in graph: image_tensor:0 , type: DT_UINT8) Outputs: 1 of 6: detection_classes (Node name in graph: detection_classes:0 , type: DT_FLOAT) 2 of 6: num_detections (Node name in graph: num_detections:0 , type: DT_FLOAT) 3 of 6: detection_boxes (Node name in graph: detection_boxes:0 , type: DT_FLOAT) 4 of 6: raw_detection_boxes (Node name in graph: raw_detection_boxes:0, type: DT_FLOAT) 5 of 6: detection_scores (Node name in graph: detection_scores:0 , type: DT_FLOAT) 6 of 6: raw_detection_scores (Node name in graph: raw_detection_scores:0, type: DT_FLOAT)